This is the complete documentation for OpsIQ. Use it to set up the product, run support, configure AI, manage your CRM, install the widget, connect sites, build connectors, call the REST API, receive webhooks, package marketplace apps, and operate self-hosted deployments. Every feature is covered with detailed examples so you can operate OpsIQ without contacting support.
Everything OpsIQ does, in one place: how each screen works, how to connect your own systems, and how to build on top of it. Start anywhere — the sections follow the same order as the admin sidebar.
Every screen in the admin: analytics, live chat, tickets, CRM, knowledge base, Site Intelligence, promotions and the customer portal. What each page does, what the numbers mean, and what to do when they look wrong.
The website widget, email mailboxes, inbound and outbound webhooks, and the connector library for platforms like Shopify, WooCommerce, WHMCS, Zendesk, Stripe, Slack and Gmail.
The Connector Builder for no-code connectors, Actions & Triggers if you would rather not use connectors at all, and the REST API with every action, scope and error code.
Roles and permissions, security, cron, diagnostics, backups, and both hosted and self-hosted deployment.
Read the screen sections in sidebar order. Each opens with what the page is for, so you can stop as soon as you have what you need.
Add Settings, Security, Diagnostics, Cron and Production. These are the sections that decide whether everything else keeps working.
Start at Connectors, then the Connector Builder, then the REST API and webhooks. Read those in order and you will have seen every way OpsIQ can talk to another system.
OpsIQ is a complete AI-first business operating platform. It replaces your analytics, live chat, helpdesk, CRM, SEO toolkit, integration middleware, and developer platform with a single unified admin where every tool shares the same customer context. Instead of stitching together 8-12 separate SaaS products, you install OpsIQ once and operate everything from one screen.
Every page load, click, scroll, and navigation event is captured in real time. See who is on your site right now, what pages they are viewing, how long they spend, and where they came from — all without third-party cookies.
Full history of every visitor with device, browser, operating system, country, city, referrer, UTM parameters, landing page, and identity resolution. Sessions group page views into journeys so you can see the entire path a visitor takes.
Timeline-style playback of captured visitor sessions. See exactly what users did — mouse movements, clicks, scrolls, page transitions, and form interactions. Use it to diagnose UX issues, understand drop-off points, and verify conversion flows.
A real-time stream of visitor activity as it happens. See page views, chat starts, form submissions, and intent signals updating every few seconds. Filter by country, device, or referrer to focus on specific traffic segments.
Pages ranked by visits, engagement time, bounce rate, conversions, and support activity. Identify your highest-performing content and pages that need improvement.
Country, city, and region breakdowns with maps. Device and browser distribution, traffic source analysis, search engine referrer tracking, and market penetration insights.
Track visitor intent signals — pricing page visits, repeated returns, documentation browsing patterns, and support interactions — to identify visitors who are likely to buy, churn, or need help.
Run split tests on your website directly from OpsIQ. Create variants, set traffic allocation, define conversion goals, and measure statistical significance. No external testing tool needed.
Capture client-side JavaScript errors from your website automatically. See error messages, stack traces, affected browsers, frequency counts, and which pages trigger them. Debug frontend issues without separate error monitoring.
Every tracked event — page views, custom events, chat interactions, form submissions, purchases — in a searchable, filterable log with full metadata.
Track orders, revenue, refunds, subscription renewals, and attribute conversions to traffic sources, campaigns, and visitor journeys. Revenue dashboards with period comparison and trend analysis.
High-intent visitors identified by behavior patterns. Score leads based on page visits, time on site, return frequency, support interactions, and custom signals. Feed leads directly into the CRM pipeline.
Automated crawling of your entire website to discover pages, detect broken links, find missing meta tags, check image alt text, validate canonical URLs, and map your site structure. Schedule crawls to run daily, weekly, or on demand.
Comprehensive technical analysis: page speed scores, mobile responsiveness, Core Web Vitals, SSL certificate status, redirect chains, duplicate content detection, crawl depth analysis, and structured data validation.
Evaluate every page for content quality, word count, readability score, keyword density, heading structure, internal/external link ratio, and freshness. Identify thin content, duplicate pages, and content gaps.
Semantic search across your entire site content. Ask natural language questions about your content and get AI-synthesized answers with source page references.
Monitor your search engine rankings for target keywords across Google and other engines. Track position changes over time, compare against competitors, and see which pages rank for which terms.
Discover keyword opportunities based on your content, competitor analysis, and search volume data. Get suggestions for new content topics, long-tail variations, and content optimization targets.
Monitor your inbound link profile. See which sites link to you, track new and lost links, analyze anchor text distribution, and identify link-building opportunities.
For businesses with physical locations: visualize your local search visibility across a geographic grid. See how your rankings vary by location, distance from your business, and search intent.
Connect your Google Business Profile to OpsIQ. Sync reviews, posts, Q&A, photos, and business information. Manage your GBP listing from inside OpsIQ. Respond to reviews, publish posts, and monitor insights.
Automated SEO and analytics reports generated on your schedule (daily, weekly, monthly). Reports include ranking changes, traffic trends, technical issues found, and content performance. White-label output available for agencies.
AI answers customer questions on your website 24/7 using your knowledge base, connected platform data (orders, accounts, subscriptions), and custom training prompts. Supports multiple AI providers: Claude (Anthropic), GPT-4o (OpenAI), Gemini (Google), and Grok (xAI).
An AI copilot inside your admin panel that helps you write ticket replies, draft customer communications, analyze support trends, look up customer data across connected platforms, and execute actions through the connector system.
The AI is grounded in your knowledge base articles, FAQ entries, product documentation, and connected platform data. It does not hallucinate — it answers from your actual business information or escalates when uncertain.
When the AI cannot resolve a question, detects frustration, or the customer asks for a human, it automatically hands off to your team with full conversation context. Configurable decline/timeout auto-revert back to AI.
Write separate instruction prompts for customer-facing AI and admin-facing AI. Control tone, boundaries, escalation rules, prohibited topics, and business-specific behavior. Per-workspace configuration supported.
Every AI conversation is logged with full request/response details, token usage, cost tracking, provider used, and model version. Review conversations to identify training gaps and improve AI accuracy.
Automated analysis of AI conversation patterns. See common questions, resolution rates, handoff frequency, customer satisfaction by topic, and areas where the AI needs better training.
Each connected site/workspace can use its own AI provider, model, temperature, token budget, and training prompts. A SaaS company can use Claude for the main product and GPT-4o for a secondary brand.
Two AI modes: Managed AI (included in your plan, no API key needed) or Bring Your Own Key (use your own OpenAI/Anthropic/Google/xAI API key for full control over model selection and costs). Both are first-class — every feature works with either mode.
Set monthly token budgets, per-conversation limits, and low-balance alerts. Track spending per provider, per workspace, and per conversation. Prevent runaway costs with hard ceilings.
Full helpdesk with departments, priorities (low/medium/high/urgent), statuses, SLA tracking, internal notes, customer replies, file attachments, and threaded conversations. No external helpdesk needed.
AI automatically triages incoming tickets by department and priority. Enable auto-reply for safe departments — the AI drafts and sends a response after a configurable delay (default: 2 minutes) so humans can intervene first.
Connect a mailbox (Gmail, IMAP, or provider-specific) to automatically create tickets from inbound emails. Replies to ticket notification emails update the ticket thread.
Embed a customer-facing ticket portal on your website. Customers can submit tickets, view their history, reply to open tickets, and check status — all branded to your design.
Merge duplicate tickets into a single thread. Merged tickets show [Merged] on the subject line. All conversation history is preserved in the surviving ticket.
After a human agent takes over a ticket, use "Return to AI" to let the AI resume auto-reply on that ticket. Useful when the human-required portion is resolved but the customer has follow-up questions the AI can handle.
AI-assisted ticket management from the inbox view. The copilot suggests replies, summarizes conversation history, recommends priority/department assignments, and can execute actions on connected platforms.
Real-time customer chat managed from a unified inbox. See all active conversations, customer identity, conversation history, and AI suggestions. Multiple agents can collaborate on a conversation.
Trigger chat messages based on visitor behavior: time on page, specific URL visited, scroll depth, return visit count, exit intent, or custom events. Example: show a help message after 60 seconds on the pricing page.
A customizable chat widget embedded on your website. Supports AI chat, live agent chat, ticket submission, knowledge base search, and visitor identity. Responsive, accessible, and brandable.
Collect customer satisfaction ratings after chat and ticket interactions. Design surveys with your own questions, rating scales, and follow-up prompts. Results feed into team performance and AI insights.
Separate AI writing assistant for composing email drafts, ticket replies, announcements, and customer communications. Uses its own prompt (not the chat brain) for professional business writing.
A single screen showing your entire CRM state: pipeline value, deal velocity, upcoming tasks, at-risk deals, agent activity, and AI recommendations. The CRM home page is your daily starting point.
Manage customer contacts with full profiles: name, email, phone, company, tags, lifecycle stage, conversation history, ticket history, purchase history, and custom fields. Link contacts to companies for B2B relationship management.
Visual Kanban-style deal board with customizable stages. Drag deals between stages, set values, assign owners, track close dates, and see pipeline value at each stage. Multiple pipelines supported.
The AI analyzes each deal and provides coaching: risk signals (gone quiet, competitor mentioned, budget concerns), next-best-action recommendations, win probability, and suggested follow-up timing.
Revenue forecasting based on pipeline data, historical close rates, and deal velocity. Reports include pipeline by stage, by owner, by source, conversion rates, average deal size, and sales cycle length.
AI-powered prospecting that identifies potential customers from your visitor data, lead signals, and connected platform activity. Suggests outreach targets, drafts initial messages, and scores lead quality.
Automated multi-step follow-up sequences: email, wait, email, check response, escalate. Define sequences for new leads, stalled deals, post-purchase check-ins, and re-engagement campaigns.
Track customer lifecycle stages (lead, prospect, customer, churned) with automated transitions. The retention agent monitors for churn signals and suggests interventions.
AI agent that continuously cleans and enriches your CRM data: deduplicates contacts, fills missing fields, validates emails, standardizes company names, and flags stale records.
Natural language CRM configuration: tell the AI what your sales process looks like and it builds your pipeline stages, custom fields, deal templates, and automation rules.
Automated audit of your CRM data quality: duplicate contacts, deals without next steps, stale pipelines, missing contact information, and data completeness scoring.
Three-level autonomy control for each CRM AI agent: Autopilot (AI acts independently), Copilot (AI suggests, human approves), and Manual (AI observes only). Set the dial per agent, per action type, and per deal value threshold.
The CRM runs four specialized AI agents: Steward (data quality), SDR (prospecting), Retention (churn prevention), and Analyst (reporting and insights). Each agent has its own Trust Dial setting.
Ready-to-use connectors for Shopify, WooCommerce, BigCommerce, Magento 2, PrestaShop, OpenCart, osCommerce, WordPress, WHMCS, Zendesk, Stripe, Amazon SES, Postmark, Resend, SendGrid, Mailgun, Gmail, IMAP, Slack, and Google Business Profile. Each connector brings platform-specific AI actions, knowledge, and data sync.
Build custom connectors for any platform — no code, AI-assisted, or full code. The builder supports REST and GraphQL APIs, OAuth authorization code flow, webhook auto-registration with Stripe and simple signature schemes, pagination (offset, page, cursor, and link-header), list/error transforms, multi-action context, and marketplace packaging.
The ConnectorBuilderAI helps you build connectors: describe what you want and it suggests actions, writes request handlers, and maps API responses. Handles authentication, pagination, and error handling automatically.
Generic in-app OAuth authorization flow that works with any connector. PKCE support, site-scoped state management, automatic token refresh. Connectors define their OAuth parameters and OpsIQ handles the flow.
Connectors don't just add platform actions — they can power core features through capability "buses". A connector declares a capability (calendar, enrichment, esign in CRM; local_listing, rank_data in Site Intelligence; web_analytics in Analytics; payment_provider in Payments; email_sync in Mailbox; video_meeting in Comms) and implements its small method set; OpsIQ discovers it via CapabilityRegistry and wires it into the UI, the Customer 360 timeline, and the AI actions. Several connectors can offer the same capability with no lock-in. See connectors/PLATFORM_CONNECTORS.md.
Connectors live in one of two tiers. Pre-installed connectors (connectors/) are registry-active out of the box — kept lean to just the keyless reference and the most-used provider per capability (e.g. Google Calendar, Gmail, Google Business Profile). Marketplace connectors (marketplace_connectors/) are discoverable but inert until installed — e.g. Microsoft 365 Calendar, Outlook, DocuSign, Stripe, GA4, SerpApi, Zoom. Every bridge surfaces a marketplace list so the UI and the AI can prompt "install from the marketplace" (notify-to-install) when a capability has no ready provider — never a broken button.
Publish your connectors to the OpsIQ marketplace. Package connectors with profile.json, knowledge.json, workflow_recipes.json, and signed archives. Marketplace supports free and paid connectors with multi-source billing.
All connectors are cryptographically signed for integrity verification. Signing is checked at install time. Use the resign_all_connectors tool to re-sign after updates.
Declarative fuzzy-matching system for connector actions. Define lookup rules and OpsIQ matches user queries to platform entities using fuzzyScore and smartRank algorithms. Example: find_country("germ") matches Germany with 88.9% confidence.
Connectors declare actions with typed parameters, authentication requirements, pagination rules, and response transforms. Over 1,200 HTTP actions across all connectors, all executable through the unified ActionExecutor.
Pre-built workflow recipes that combine multiple connector actions into common business workflows. Example: "When a new order comes in on Shopify, create a ticket, update the CRM, and send a Slack notification."
Connectors declare knowledge topics that sync to the AI knowledge base. Platform-specific help articles, setup guides, and troubleshooting steps are automatically available to the AI without manual knowledge entry.
Full API for visitors, events, tickets, contacts, settings, and connector operations. API key authentication, rate limiting, and JSON request/response format. OpenAPI specification available.
Outbound webhooks for all major events: orders, invoices, subscriptions, tickets, chats, leads, user registrations, and cart abandonment. Async delivery via job queue, exponential backoff retry, SSRF protection, and auto-disable after 15 consecutive failures.
Every outbound webhook is signed with HMAC-SHA256. Verify signatures on your receiving end to ensure webhook authenticity. Stable event_id for idempotent processing.
JavaScript SDK for the tracking widget with identity resolution, custom event tracking, programmatic chat control, and page-specific configuration. Install recipes for React, Vue, Next.js, WordPress, Shopify, and static HTML.
Server-side SDKs and plugins for common frameworks. WordPress plugin, WHMCS module, and generic PHP/Node.js integration libraries.
Push custom events into OpsIQ from your backend: purchases, signups, feature usage, errors, or any business event. Events appear in the visitor timeline and feed into analytics and AI context.
Build automated workflows triggered by events: new ticket, chat started, visitor identified, deal stage changed, or custom event. Triggers execute connector actions, send webhooks, update records, or notify your team.
Visual API and webhook builder inside the admin. Configure endpoints, test requests, inspect responses, and debug integrations without leaving OpsIQ.
Three roles: Owner (full access, billing, danger zone), Full Admin (all operations except billing and danger zone), and Agent (only assigned departments, no settings access). All sensitive pages are hard-gated by role.
Owner and Full Admin-only dashboard showing resolution times, workload distribution, handle depth, channel breakdown, trend analysis, SLA compliance, CSAT by agent, reopen rates, team rollups, and busiest hours. Updated in real time.
Real-time who-is-online widget in the admin navigation. Heartbeat-based presence detection (60-second ping, 300-second extend, auto-prune after 120 days). See which admins are currently active and on which pages.
Each connected site (workspace) can have its own AI configuration, connector settings, ticket departments, branding, and operational rules. Global defaults cascade to workspaces that do not override them.
Manage multiple websites from one OpsIQ admin. Each site has its own tracking snippet, visitor data, and configuration. Switch between sites from the admin navigation.
Registry-driven email templates for all automated communications: ticket notifications, chat transcripts, welcome emails, password resets, and system alerts. Edit HTML templates with live preview, variable chips ({{customer_name}}, {{ticket_id}}), and global layout control.
Import contacts, tickets, and knowledge articles from CSV/JSON. Export visitor data, analytics reports, and CRM records. Data portability for migration between systems.
OIDC (OpenID Connect) single sign-on with full runtime: discovery, authorize, callback, JWKS signature verification, userinfo, domain allow-listing, JIT provisioning, and status guards. SAML configuration support.
Built-in DSAR (Data Subject Access Request) handling: export and delete customer data across all systems. Contact-level data erasure covering sessions, tickets, chats, CRM records, and email logs. Consent management and retention policies.
Block specific IPs or ranges from accessing your site or admin. Automatic lockout after configurable failed login attempts. Failed login log with IP, timestamp, and user agent.
HTTP Strict Transport Security, Permissions-Policy headers, and secure cookie configuration. Self-hosted deployments can add CSP headers through server configuration.
SSRF guard blocks webhooks to private/metadata IP ranges. Async delivery prevents slow endpoints from blocking your application. Auto-disable after 15 consecutive failures prevents resource waste on dead endpoints.
Strict data isolation between workspaces. Visitors, tickets, chats, CRM records, and AI conversations are scoped to their workspace. Cross-workspace data leaks are prevented at the query layer.
Support for monthly, quarterly, semi-annual, yearly, biennial, and lifetime billing cycles. Currency normalization for multi-currency gateways. FX conversion on payment receipt for accurate revenue reporting.
Built-in cron system for scheduled tasks: report generation, data cleanup, subscription renewals, AI training updates, connector syncs, and webhook retries. Health diagnostics with warning indicators.
Configure how long visitor sessions, chat transcripts, AI conversation history, and event logs are retained. Automatic cleanup of aged data to manage storage and comply with retention policies.
System health dashboard showing cron status, database connectivity, AI provider status, connector health, webhook delivery rates, and storage usage. Warnings for configuration issues and performance bottlenecks.
Track every visitor from landing to purchase. Let AI answer "Where is my order?" using live Shopify/WooCommerce/BigCommerce data. Manage support tickets with AI auto-reply. Run a deal pipeline for wholesale and B2B leads. Attribute revenue to traffic sources and campaigns. Monitor cart abandonment with recovery workflows.
Monitor trial signups and feature adoption with session replay. Track which documentation pages convert. Automate support with AI grounded in your API docs. Manage subscription revenue and expansion deals in the CRM. Connect your billing platform (Stripe, WHMCS) for real-time MRR tracking. Use the CRM lifecycle agent to detect and prevent churn.
Offer white-label analytics and support portals to clients. Track leads across multiple client sites from one admin. Use the CRM pipeline to manage prospects and project leads. Build custom connectors for client platforms using the Connector Builder. Generate branded SEO reports with Site Intelligence. Resell marketplace connectors as value-add services.
Deep WHMCS integration with 80+ AI actions for order lookup, service management, DNS, and provisioning. Auto-reply to common hosting questions using knowledge base articles. Track server status page engagement. Manage technical support tickets with AI triage. Use the billing cycle ladder for monthly through lifetime hosting plans.
Deflect 40-70% of repetitive questions with AI chat grounded in your knowledge base. Route tickets to the right department with AI triage. Track team performance with resolution times, SLA compliance, and CSAT by agent. Use the writing AI for professional reply drafting. Monitor support trends with AI insights.
Full Site Intelligence suite: crawls, technical SEO audits, content analysis, SERP rank tracking, keyword research, backlink monitoring, and local SEO grid. Google Business Profile management. Scheduled white-label reports for clients. A/B testing built into the tracking widget. Campaign attribution across all channels.
Start with just tracking and chat — free to add. Then add tickets when you get support volume. Turn on the CRM when you start selling. Connect platforms as you adopt them. OpsIQ scales from one person doing everything to a 50-person operations team with RBAC, departments, and per-agent performance tracking.
Enterprise SSO with OIDC for centralized authentication. Multi-workspace isolation for separate brands or divisions. Per-workspace AI configuration with different providers and models. Custom connector development with marketplace distribution. Audit trails, GDPR compliance tools, and data retention policies.
Full website analytics with visitor tracking, session replay, page rankings, geo intelligence, intent funnels, conversion attribution, events, A/B experiments, and JS error tracking — all privacy-friendly, all in one platform.
AI-powered customer chat with knowledge grounding, multi-provider AI support, auto-handoff to humans, proactive chat rules, conversation history, CSAT surveys, and ticket creation from chat.
Native ticket system with departments, priorities, SLA, AI triage, auto-reply with configurable delay, email ingestion, ticket embed, ticket merging, inbox copilot, and Return to AI capability.
AI-first CRM with four specialized AI agents (Steward, SDR, Retention, Analyst), Trust Dial autonomy control, pipeline board, deal coaching, forecasting, sequences, lifecycle management, and natural language CRM configuration.
Site Intelligence with automated crawls, technical SEO audits, content analysis, AI search, SERP rank tracking, keyword research, backlink analysis, local SEO grid, and scheduled white-label reports.
Connector SDK with 25+ pre-built connectors, Connector Builder (no-code/AI-assisted/full-code), OAuth flow, webhook auto-registration, a large catalogue of connector action definitions that become executable only when the relevant connector is installed, enabled, configured and validated, SmartLookup, marketplace, and signed packages.
Full GBP connector: review management, post publishing, Q&A management, photo uploads, business info sync, insights tracking, and AI-powered review responses — all from inside OpsIQ.
Registry-driven email templates with visual editor, live preview, variable chips, global layout control, and per-event customization for all automated communications.
Built-in team performance: resolution times, workload, SLA compliance, CSAT by agent, reopen rates, busiest hours, real-time presence tracking, and per-department breakdowns.
Native DSAR handling, data erasure across all systems, consent management, retention policies, and audit trails — built into the platform, not bolted on.
OpsIQ runs on managed infrastructure. Product operation is the same; infrastructure, updates and backups are handled by the hosting operator. This handbook documents the OpsIQ runtime rather than marketing or commercial account flows.
Install OpsIQ on your own server for full control over data, branding, updates, and integrations. Requires PHP 8.4+, MySQL 5.7+ or MariaDB 10.3+, HTTPS, and the OpsIQ unified cron. Supports WHMCS integration as an addon module. Full data sovereignty.
Follow the 30-minute setup guide to get tracking, chat, and basic AI working on your website.
Learn how to read your visitor data, set up session replay, identify leads, and understand traffic patterns.
Run your first SEO crawl, check your technical health, and set up SERP rank tracking for your target keywords.
Set up departments, configure auto-reply with AI, connect email ingestion, and train your AI to handle common questions.
Organize contacts, build your pipeline, configure the AI agents, set Trust Dial levels, and let AI coach your deals.
Connect your e-commerce/billing platform, build custom connectors, set up webhooks, and integrate via the REST API.
Follow these steps in order to get OpsIQ running. Each step builds on the previous one. By the end, you will have visitor tracking, AI chat, and basic ticket support.
Enter your license key in the OpsIQ admin. Go to Settings, enter the licence key supplied for this installation, and click Activate. You should see a green "Active" badge.
Go to Settings and fill in your business name, industry, timezone, default currency, and support hours. This information is used by the AI when answering customer questions and by analytics for timezone-aware reports.
Go to AI Configuration and select your provider. Options: Claude (Anthropic), GPT-4o (OpenAI), Gemini (Google), or Grok (xAI). Enter your API key if using your own key, or use Managed AI if available on your plan. Click "Test connection" to verify.
Go to Connected Sites, copy the tracking snippet, and paste it before the closing </body> tag on your website. Visit your site in a private browser, then check the Live Feed in OpsIQ — you should see yourself as a visitor within 15 seconds.
If you use Shopify, WooCommerce, WHMCS, or another supported platform, go to Connectors, find your platform, enter credentials, and click Test Connection. This gives the AI access to order data, customer profiles, and support context.
Go to Team, click Invite, enter their email and role (Owner, Full Admin, or Agent). Agents can only access departments you assign them to. Owners and Full Admins see everything.
Copy the snippet from Connected Sites and paste it on your website. This gives you tracking immediately.
Go to AI Configuration, pick a provider, enter your key, and save. AI chat is now active on your website.
You now have visitor tracking and AI chat. You can add tickets, CRM, and connectors later.
Go to Knowledge Base and add 5-10 articles about your most common questions. The AI uses these to answer customers accurately.
Go to Tickets and create at least 2 departments (e.g., "Sales" and "Support"). Assign team members to each.
Go to AI Training and write a Customer AI prompt that describes your business, tone, and escalation rules.
Open your website in a private browser and chat with the AI as a customer. Check that answers are accurate and tone is right.
Set up a mailbox (Gmail, IMAP, or a provider connector) so customer emails automatically create tickets.
Enable ticket auto-reply for safe departments. Set a 2-minute delay so humans can intervene first.
Check Dashboard daily. Look at visitor count, bounce rate, top pages, and sales if tracking is connected.
Open AI CRM and let OpsIQ import contacts from your connected platforms. Review the pipeline board.
Check AI History weekly. Look for wrong answers, missed questions, and opportunities to add knowledge.
Create a rule: if a visitor is on the pricing page for 60+ seconds, show a chat message asking if they need help.
If you use a platform that does not have a built-in connector, use the Connector Builder to create one.
Connect OpsIQ to your internal tools using outbound webhooks for ticket, chat, and sales events.
Yes. The AI provider is optional. OpsIQ works as a pure analytics and ticketing platform without AI. You can enable AI later when you are ready.
Yes. Connectors enrich the AI with platform-specific data, but OpsIQ works without them. Start with tracking and chat, then connect platforms when needed.
You can still use OpsIQ for tickets, CRM, and team collaboration. The tracking widget is optional. Add it when your website is ready.
Visit your website in a private/incognito browser. Then check OpsIQ Live Feed. You should see your visit appear within 15 seconds. If not, check: (1) is the snippet pasted correctly, (2) is the site key correct, (3) are ad blockers or CSP headers blocking the script.
OpsIQ is not one feature. It is an operator admin, a website widget, an AI support layer, a ticket/helpdesk system, a full CRM with four AI agents, a Site Intelligence suite, a connector platform with marketplace, and a developer surface. Here is every component it contains.
Dashboard, live feed, visitor log, sessions, session replay, top pages, geo intelligence, intent funnel, events log, A/B experiments, JS error tracking, sales and conversions, leads, AI CRM (command center, contacts, companies, deals, pipeline board, deal coaching, forecast, prospecting, sequences, lifecycle, data steward, Build my CRM, CRM config, CRM health, CRM API), tickets, chat inbox, proactive rules, CSAT surveys, email drafts, mailboxes, AI configuration, AI training, AI history, AI insights, knowledge base, connectors, Connector Builder, marketplace, connected sites, team management, team performance, settings, diagnostics, and license.
Website tracking widget (analytics beacon + chat + ticket embed + knowledge base search), client AI chat, ticket submission portal, visitor identity token handling, file uploads, CSAT survey widget, and customer-facing help center.
Command center, contacts, companies, deals, pipeline board with drag-and-drop, AI deal coaching (risk signals, next-best-action, win probability), revenue forecasting, prospecting agent (SDR), automated follow-up sequences, lifecycle and retention management, data steward agent (dedup, enrich, validate), Build my CRM (natural language setup), CRM configuration, CRM health check, Trust Dial (Autopilot/Copilot/Manual per agent), four specialized agents (Steward, SDR, Retention, Analyst), and CRM API with webhooks.
Automated site crawls, technical SEO audit (speed, mobile, Core Web Vitals, SSL, redirects, structured data), content audit (word count, readability, keyword density, heading structure), AI-powered semantic search, SERP rank tracking across engines, keyword research with opportunity scoring, backlink analysis and monitoring, local SEO grid visualization, Google Business Profile sync (reviews, posts, Q&A, photos, insights), scheduled reports (daily/weekly/monthly), and white-label PDF output for agencies.
Customer AI chat (multi-provider: Claude, GPT-4o, Gemini, Grok), admin AI assistant, AI knowledge grounding, auto-handoff with configurable revert, per-workspace AI configuration, managed AI and BYOK modes, token budgets and cost controls, AI history audit trail, AI insights analysis, ticket AI triage, ticket auto-reply with delay, writing AI (separate prompt), and admin inbox copilot.
25+ pre-built connectors (Shopify, WooCommerce, BigCommerce, Magento 2, PrestaShop, OpenCart, osCommerce, WordPress, WHMCS, Zendesk, Stripe, Amazon SES, Postmark, Resend, SendGrid, Mailgun, Gmail, IMAP, Slack, Google Business Profile), Connector Builder (no-code/AI-assisted/full-code), generic OAuth flow with PKCE, connector marketplace with paid distribution, connector signing and verification, SmartLookup fuzzy matching, a large catalogue of HTTP action definitions whose availability depends on connector installation, enablement, configuration and validation, knowledge sync, and workflow recipes.
REST API (visitors, events, tickets, contacts, settings, connectors), OpenAPI spec, API key management, outbound webhooks (async, signed, retried, SSRF-guarded, auto-disable), Events API for custom event ingestion, widget SDK with identity resolution, triggers cookbook, action contracts, connector SDK, marketplace packaging, server-side SDKs (PHP, Node.js), WordPress plugin, and WHMCS module.
Business identity, AI configuration per workspace, tracking settings, security (HSTS, Permissions-Policy, IP blocking, brute-force protection), RBAC (Owner/Full Admin/Agent), team management and invitations, team performance dashboard, admin presence tracking, email template system (registry-driven, visual editor, variable chips), survey designer, domain knowledge, ticket embed configuration, branding, data retention policies, billing cycle management (monthly through lifetime), currency normalization, import/export, diagnostics, cron automation, and danger zone.
Enterprise SSO (OIDC with JWKS verification, JIT provisioning), SAML configuration, GDPR/DSAR handling (export and delete across all systems), data erasure per contact, consent management, workspace isolation (strict query-layer scoping), audit trails, failed login tracking, IP blocking, outbound webhook SSRF guard, and data retention policies.
Managed AI proxied through OpsIQ provider URL, Site Intelligence provider calls, license validation, health reports, PDF rendering, and data exports — all routed through the configured provider endpoint, not direct vendor API calls from your server.
Daily command center: traffic overview, live visitor count, sales summary, support queue depth, AI activity alerts, and quick-action shortcuts.
Real-time visitor stream: page views, chat starts, form submissions, and intent signals updating every few seconds with country, device, and referrer filters.
Complete visit records: device, browser, OS, country, city, referrer, UTM params, landing page, identity, and full page-view history.
Journey-grouped view: page flow, time on page, scroll depth, conversion path, and exit page for each visitor session.
Timeline playback of captured events: mouse movement, clicks, scrolls, page transitions, and form interactions for UX diagnosis.
Page performance ranking: visits, engagement time, bounce rate, conversions, support interactions, and trend direction.
Geographic reports: country/city breakdown with maps, device distribution, browser share, traffic sources, and market penetration.
Behavioral intent tracking: pricing visits, return frequency, doc browsing, support interactions, and custom signals with conversion probability.
Searchable event history: page views, custom events, chats, purchases, form submissions with full metadata and filtering.
Split testing: create variants, allocate traffic, define goals, measure significance, and declare winners — all from inside OpsIQ.
Frontend error capture: messages, stack traces, affected browsers, frequency, affected pages, and trend tracking.
Revenue dashboard: orders, renewals, refunds, attribution by source, campaign performance, period comparison, and trend analysis.
Lead management: scoring, source attribution, CRM status, contact details, behavioral signals, and pipeline routing.
Full CRM: command center, contacts, companies, deals, pipeline board, coaching, forecast, prospecting, sequences, lifecycle, data steward, and health.
Support inbox: departments, priorities, SLA tracking, AI triage, auto-reply, internal notes, file attachments, merging, and status management.
Live conversations: AI and human chat, handoff, collaboration, conversation history, customer identity, and CSAT collection.
Email management: inbound mail to tickets, AI-assisted reply composition, template selection, and mailbox configuration.
Provider setup: select Claude/GPT-4o/Gemini/Grok, enter API key or enable managed AI, set model, temperature, and token budgets per workspace.
Prompt authoring: customer-facing and admin-facing prompts, escalation rules, tone guidelines, prohibited topics, and boundary configuration.
Conversation audit: full request/response logs, token usage, cost per conversation, provider, model version, and resolution outcome.
Pattern analysis: common questions, resolution rates, handoff frequency, satisfaction by topic, and training gap identification.
AI training content: articles, FAQ entries, product docs, connector knowledge, and custom entries. Semantic and keyword search for AI grounding.
Integration hub: installed connectors, health status, action inventory, credential management, marketplace browser, and Connector Builder.
Site management: tracking snippet, site key, widget configuration, domain settings, and per-site workspace isolation.
User management: invite members, assign roles (Owner/Full Admin/Agent), set department access, and manage permissions.
Ops metrics: resolution times, workload, SLA compliance, CSAT by agent, reopen rates, busiest hours, and trend analysis.
Configuration: business identity, AI setup, tracking, security, branding, retention, email templates, billing, import/export, diagnostics, and danger zone.
Plan management: license status, plan features, usage metrics, renewal date, and upgrade/downgrade options.
The onboarding journey turns a new installation into a working OpsIQ workspace and can be left and resumed without losing progress.
The journey stores completion per workspace. Returning to Onboarding should show completed steps rather than forcing a restart.
The Analytics hub brings support, chat, AI, visitor and agent reporting into one date-filtered workspace with exportable results.
Use Analytics Hub for cross-team review; use the dedicated Visitors, Tickets, CSAT, AI History and Team Performance pages when you need record-level investigation.
The Dashboard is your daily command center. It shows traffic, sales, support load, AI activity, and conversion indicators in one view. Open it every morning to understand what happened overnight and what needs attention today.
The number of people on your website right now. This updates every 15 seconds (configurable). A visitor is "active" if they loaded a page within the active threshold (default: 90 seconds).
How many different people visited your website in the selected date range. OpsIQ identifies visitors by a combination of browser fingerprint, cookies, and (if available) authenticated identity.
New visitors are seeing your site for the first time. Returning visitors have been seen before. A high returning ratio means your content brings people back. A very low returning ratio might mean you are not retaining interest.
The percentage of visitors who left after viewing only one page. A bounce rate above 70% on landing pages usually means the page is not meeting visitor expectations — wrong content, slow load, or poor mobile experience.
How long visitors spend on each page on average. Very short times (under 10 seconds) on content pages suggest the content is not engaging. Very long times on checkout pages might indicate confusion.
IP addresses that OpsIQ identified as potentially malicious based on threat scoring, brute force attempts, or manual blocks. Click to see the blocked IPs list.
Revenue from connected sales sources (Shopify, WooCommerce, Stripe, etc.). If sales show zero, check that your platform connector is connected and syncing.
A line graph showing visitor count over the selected date range. Look for patterns: traffic spikes (campaigns, press mentions), dips (weekends, holidays), and sustained trends.
Where your visitors come from: direct, Google, social media, email campaigns, referral sites. Use this to understand which marketing channels are working.
Chrome, Safari, Firefox, Edge, and others. If a browser has unusually high bounce rate, you might have a CSS or JavaScript compatibility issue.
A grid showing visitor activity by hour of day and day of week. Use this to decide when to schedule support shifts, when to send announcements, and when to run promotions.
The last few visitors with country, page, device, and source. Useful for quick spot-checks during campaigns.
Click the date picker at the top right of the Dashboard. Presets: Today, Yesterday, Last 7 days, Last 30 days, This month, Last month, Custom range. All dashboard cards and charts update when you change the range. The date range is in your configured timezone (set in Settings).
Your bounce rate jumped from 45% to 78% this week. What happened?
Check: (1) Did a specific landing page change? Go to Top Pages and sort by bounce rate. (2) Did a campaign send traffic to the wrong page? Check Top Sources for new referrers. (3) Did the site break on mobile? Check browser breakdown for mobile bounce rate. (4) Is it bot traffic? Check the Threats counter and visitor details for suspicious patterns.
The dashboard shows 0 live visitors even though your website is getting traffic.
Check: (1) Is the tracking widget installed? View your website source code and search for the OpsIQ script. (2) Is the site key correct? Compare the data-site-key in the script with the key in Connected Sites. (3) Is a CDN or cache serving a stale page without the script? Clear your CDN cache. (4) Is an ad blocker or CSP header blocking the script? Check the browser console for blocked requests.
Your heatmap shows heavy activity between 10am-2pm on weekdays but nothing on weekends.
This is a B2B traffic pattern. Your visitors are working professionals browsing during business hours. Implications: (1) Schedule support staff for weekday business hours. (2) Send email campaigns Tuesday-Thursday morning. (3) Weekend maintenance windows are safe. (4) Consider offering live chat only during peak hours.
You see a sudden spike of 500 visitors in one hour from an unknown referrer.
Go to Visitors and filter by the time window. Check: (1) Are they real visitors or bots? Look at session duration, pages viewed, and browser diversity. (2) If real, what page are they landing on? (3) Check the referrer URL — someone might have linked to you from Reddit, Hacker News, or a popular blog. (4) If it is bot traffic, check Threats and consider blocking the IP range.
OpsIQ counts visitors differently. It uses browser fingerprinting and cookies, while GA uses a different cookie model. Bot filtering, ad blocker rates, and script loading order can also cause differences. A 10-20% variance is normal.
The dashboard is designed for the OpsIQ admin. For external reporting, use the REST API to pull analytics data and display it in your own dashboard or BI tool.
It counts IP addresses that OpsIQ has flagged or blocked based on threat scoring, repeated failed logins, or manual blocks. Click the counter to see details and manage blocks.
The Visitor Log shows every visit to your tracked websites. Each row is one page load from one visitor. Use it to investigate individual visitor behavior, find campaign traffic, and diagnose tracking issues.
The visitor's IP address. Anonymized if IP anonymization is enabled in Settings. Click to see all visits from this IP.
Detected from the IP address using GeoIP data. Requires MaxMind GeoLite2 database or Cloudflare geo headers. Shows "Unknown" if GeoIP is not configured.
The page the visitor loaded. Click to see the full URL. Long URLs are truncated in the table view.
Where the visitor came from before landing on your site. "Direct" means no referrer (bookmarks, typed URLs, some app links).
The visitor's device type (desktop, mobile, tablet) and browser (Chrome, Safari, Firefox, etc.). Parsed from the User-Agent header.
When the page was loaded, in your configured timezone.
If the visitor is identified (via identity token or login), their name or email appears. Otherwise shows as anonymous.
Filter visits to a specific time period. Default: last 7 days.
Search by IP address, page URL, referrer, country, or customer email.
Filter by traffic source: Direct, Organic, Social, Referral, Paid, Email.
Filter by Desktop, Mobile, or Tablet.
Show only first-time visitors who have never been seen before.
You launched a Google Ads campaign with UTM parameter utm_source=google&utm_medium=cpc&utm_campaign=summer-sale. How do you find those visitors?
In the Visitor Log, use the Search field and type "summer-sale". The search checks referrer URLs and page URLs, so UTM parameters will match. You can also filter by Source = Paid to narrow down further.
A customer emailed saying "your checkout page was broken yesterday around 3pm." How do you find their session?
Search by their email address in the Visitor Log. If they were logged in, their identity will appear. Set the date range to yesterday. Look for visits to checkout pages around 3pm. Click their session to see the full page flow and any JavaScript errors.
You notice an IP address has 500+ visits in one hour, all to the same page.
This is likely a bot, scraper, or broken script. Go to the Visitor Log and search by that IP. Check: (1) Are all visits to the same page? (2) Is the user agent a known bot? (3) Is the session duration always 0 seconds? If confirmed as a bot, go to Security and block the IP.
Raw page loads. One row per page view. Best for: investigating specific pages, finding traffic from specific sources, diagnosing tracking issues.
Groups page views into journeys. One row per visit session. Best for: understanding user flow, measuring session duration, identifying conversion paths.
Real-time activity. Shows visitors currently on your site. Best for: monitoring campaigns in progress, spotting high-intent visitors, watching support opportunities.
OpsIQ tracks all visits by default. To exclude your own traffic, add your IP address to the excluded IPs list in Settings > Tracking.
GeoIP lookup requires either Cloudflare geo headers (automatic if using Cloudflare) or a MaxMind GeoLite2 database file. Check Settings > GeoIP Path to ensure the database file exists and is readable.
Visitor data is kept for the retention period configured in Settings (default: 90 days). After that, old records are automatically cleaned up by the cron retention job.
Sessions group multiple page views from the same person into a single journey. While the Visitor Log shows individual page loads, Sessions show you how people move through your website from landing to exit.
A session starts when a visitor loads their first page. It continues as they navigate to more pages. A session ends when the visitor is inactive for the session timeout period (default: 30 minutes) or closes their browser. The same person returning after the timeout creates a new session.
The landing page — where the visitor entered your site. This tells you which content is attracting people.
The exit page — where the visitor left. If the exit page is often the pricing page, visitors might be comparing you with competitors.
How many pages the visitor viewed in this session. More pages usually means more engagement, but it can also mean the visitor is lost.
Total time from first page load to last activity. Sessions with 0 seconds are bounces (single page view, no further interaction).
The traffic source for this session (Direct, Google, social, referral, etc.).
Detected country from the visitor's IP address.
Whether the visitor left after viewing only one page. Bounced sessions have a red indicator.
Whether the visitor completed a tracked conversion event (purchase, signup, form submission) during this session.
Search by IP, email, page URL, or referrer within sessions.
Filter by New visitors (first session ever) or Returning visitors.
Filter by traffic source.
Filter by country to see sessions from a specific region.
Filter to show only bounced sessions or sessions flagged as threats.
A customer just made a large purchase. You want to understand their journey before buying.
Search for their email in Sessions. Look at all their sessions: (1) How many times did they visit before buying? (2) Which pages did they view? (3) Did they use chat or open a ticket? (4) What was the total time from first visit to purchase? This tells you the typical buying journey for high-value customers.
A customer usually logs in from Nigeria but you see a session from Russia.
Search for the customer email in Sessions. Compare the country of recent sessions. If a session suddenly appears from an unusual country, it could be legitimate travel or a compromised account. Check: (1) Was there a failed login attempt before the successful one? (2) Did the IP change within the same session? Flag it for review if suspicious.
A visitor is a person (or browser). A session is one visit from that person. The same visitor can have many sessions over time. Sessions expire after the timeout period.
OpsIQ uses a combination of cookies, browser fingerprinting, and authenticated identity tokens. If a visitor clears their cookies and is not logged in, they will appear as a new visitor.
The Live Feed shows visitors currently on your website in real time. It updates automatically every 15 seconds (configurable). Use it during campaigns, product launches, and busy support periods to see what is happening right now.
The Live Feed polls your server at the configured refresh interval. Each row shows a visitor currently on your site: their current page, country, device, source, session duration, and page count. Green highlights indicate new activity since the last refresh. Click any visitor row to see their full session details.
How often the live feed updates. Default: 15 seconds. Lower values (5-10s) give faster updates but use more server resources. Higher values (30-60s) are easier on the server but less real-time.
How long a visitor is considered "active" after their last page load. Default: 90 seconds. If a visitor loads a page and then reads it for 2 minutes without clicking, they disappear from the live feed after the threshold.
Click Pause to stop auto-refresh. Useful when you want to read details without the list updating. Click Resume to restart.
You are running a product launch. A visitor just viewed your homepage, then features, then pricing, and is now on the checkout page.
This is high-intent behavior. In the Live Feed: (1) Click the visitor to see their session. (2) If they start a chat, the AI will have their full journey context. (3) If they leave without buying, they become a lead in the Leads page. (4) Consider creating a proactive rule to show a chat message when visitors reach checkout.
You see 20 visitors from the same IP address, all on different pages.
This could be: (1) An office with shared internet — check if the visits look like real humans (different pages, reasonable session durations). (2) A bot or scraper — check if sessions are very short (0-1 seconds) or hitting pages systematically. (3) A load test — if you are running automated tests, exclude the IP from tracking.
Check: (1) Is the widget script installed on your site? (2) Is the site key correct? (3) Has the active threshold passed? (4) Are visitors using aggressive ad blockers that block the tracking script?
The live feed shows all active visitors. For filtered views, use the Visitor Log or Sessions pages with their filter options. The live feed is intentionally simple for real-time monitoring.
Top Pages shows which pages on your website get the most traffic. It ranks pages by visits, engagement, conversions, and support activity. Use it to identify your best content, find underperforming pages, and discover pages that generate support questions.
The full URL path. Similar pages (e.g., product variants) are grouped if grouping is enabled.
Total number of times this page was loaded in the selected date range.
How many different people viewed this page (a person viewing the same page twice counts as 1 unique).
Percentage of visitors who left your site after viewing only this page. High bounce on landing pages = problem. High bounce on "thank you" pages = normal.
How long visitors spend on this page on average. Very short = not reading. Very long on a form page = confusion.
How many chat conversations were started from this page. High chat starts on a product page means the product description might be unclear.
How many support tickets were created by visitors who were on this page. High ticket count = potential UX or content problem.
Your support team keeps getting questions about shipping. Which page is failing?
Sort Top Pages by "Support tickets" or "Chat starts" descending. Look for product or shipping pages with high counts. The page with the most support questions probably has unclear information. Fix the page content and monitor whether support questions decrease.
You want to know where people leave your site most often.
Sort by bounce rate descending. Ignore pages that are naturally exit points (thank you pages, confirmation pages). Focus on pages that should keep visitors engaged: product pages, pricing, features. These need better calls-to-action or clearer content.
Check: (1) Is the tracking widget installed on that specific page? Single-page apps might not fire page views on route changes. (2) Is the page excluded in tracking settings? (3) Is a CDN serving a cached version without the tracking script?
Geo Intelligence shows where your visitors come from by country, city, device, and source. Use it to understand your geographic markets, decide on language support, plan regional marketing, and detect suspicious traffic patterns.
The top countries ranked by visitor count. Click a country to drill down to city-level data.
Visual breakdown of traffic by country. Hover for percentages.
Every country with visitors in the selected period, showing visitor count, session count, bounce rate, and conversion rate.
Click a country to see which cities within that country have the most visitors.
OpsIQ detects visitor location in two ways: (1) Cloudflare geo headers — if your site is behind Cloudflare, country detection is automatic and accurate. (2) MaxMind GeoLite2 database — download the free database and configure the path in Settings. City-level detection requires the GeoLite2-City database, not just the Country database.
Visitors using VPNs will appear to come from the VPN server's location, not their real location. This is by design — OpsIQ cannot and should not try to bypass VPNs. If you see unusual country patterns, consider that 10-20% of visitors may be using VPNs.
You run a US-only business but 30% of traffic is from India.
Check: (1) Are these real visitors? Look at session duration and pages viewed. Very short sessions with high bounce = likely bots. (2) Are they finding you through Google? Check the referrer. (3) Could they be potential customers in a new market? If the traffic is genuine and engaged, consider adding India-specific pricing or content.
You want to know if you should translate your site into Spanish.
Look at the Geo Intelligence page. If Spanish-speaking countries (Spain, Mexico, Colombia, Argentina, etc.) together represent more than 10% of your traffic with low bounce rates, translation is worth considering. Also check: do visitors from these countries convert at similar rates to your primary market?
You need to decide support team hours for different time zones.
Use the hourly heatmap (on Dashboard) combined with Geo Intelligence. If 40% of your traffic is from Europe and 35% from the US, you need coverage for both time zones. The heatmap shows exactly when each region is most active.
Germany shows 500 visitors this month but zero sales.
Investigate: (1) Are they bouncing on the pricing page? Maybe your pricing is not competitive for that market. (2) Do you ship to Germany? Check if your shipping page mentions Germany. (3) Is your checkout available in EUR? Currency mismatch causes abandoned checkouts. (4) Are these visitors from a single source that might be low-intent traffic?
GeoIP is not configured. Either set up Cloudflare (which adds geo headers automatically) or download the MaxMind GeoLite2 database and set the file path in Settings > GeoIP Path.
OpsIQ does not block by country. Use your web server or Cloudflare firewall rules for country-based blocking. OpsIQ focuses on visibility, not access control at the network level.
The Intent Funnel classifies visitor sessions by what they seem to be trying to do: browse, research, compare, buy, get support, or leave. Use it to understand the balance of intent on your site and identify where visitors drop off.
Visitors casually looking around. Short sessions, few pages, no specific goal detected.
Visitors reading product/feature pages, docs, or blog content. Longer time on page, multiple content pages.
Visitors viewing pricing, plan comparison, or feature tables. Often returning visitors.
Visitors on checkout, cart, or signup pages. High intent — these are your conversion opportunities.
Visitors on help pages, contact pages, or starting chat. They need assistance.
Visitors who left after one page. Low engagement.
1,000 visitors viewed Pricing. 270 opened chat. 92 started checkout. 64 paid. 18 opened support tickets.
Pricing page is strong enough to create intent. Chat is important in the buying path — 27% of pricing visitors engage with chat before buying. Support tickets after checkout may indicate activation or payment confusion. Action: (1) Ensure AI chat answers pricing questions well. (2) Investigate the 18 post-checkout support tickets for common issues.
Use the date range filter to compare intent patterns across periods. For example, compare Black Friday week to a normal week — you should see a spike in "Buying" intent and possibly "Support-seeking" if your promotions created confusion.
OpsIQ looks at page URLs, session behavior, and time patterns. Pricing pages signal comparison intent. Checkout pages signal buying intent. Help/contact pages signal support intent. The classification is heuristic-based, not AI-driven.
Not currently. The categories are built-in based on common web behavior patterns. Custom intent classification is on the roadmap.
Build custom journeys from page or event steps, measure cohort retention and discover the paths visitors most often take through a site.
Create a named funnel with 2 to 8 page or event steps. Set the conversion window in hours; a visitor counts only when the ordered steps occur inside that window. Use the drop between steps to choose where to inspect replay, errors, page speed or messaging.
Retention groups visitors into weekly cohorts and shows how many return. Choose a 6, 8 or 12 week horizon. Keep the cohort definition stable when comparing releases or campaigns.
Paths lists the most common page sequences. Choose a path length from 2 to 5 pages to keep the result readable. Use it to find unexpected detours, common exits and content that repeatedly precedes conversion.
Custom funnels are analytical definitions, not automations. To react to a completed step, create a conversion goal, segment, trigger or workflow.
Session Replay lets you watch a timeline of what a visitor did on your site: which elements they clicked, how they scrolled, where they hovered, and what they typed (with sensitive fields masked). Use it to diagnose UX problems, verify bug reports, and understand customer confusion.
From the Sessions page, look for sessions with reported issues, high page counts with no conversion, or sessions from customers who filed support tickets. Click the replay icon to open the timeline view.
A horizontal bar showing events over time. Click anywhere on the timeline to jump to that moment.
Each page load appears as a segment on the timeline. You can see the exact path the visitor took.
Click events are shown as dots on the timeline and as highlights on the page reconstruction.
Scroll depth is tracked so you can see how far down the visitor read.
Field focus and typing events are recorded. Text input in password and sensitive fields is masked by default.
Session replay masks sensitive form fields by default. Password fields, credit card inputs, and fields marked with data-opsiq-mask are never recorded. You can configure additional masking rules in Settings > Tracking. If your compliance policy requires it, you can disable replay entirely.
A customer says "I tried to buy but the button did nothing."
Find their session in the Sessions page (search by email). Open the replay. Watch the checkout page segment. Look for: (1) Did they click the right button? (2) Did a JavaScript error occur? (Check the JS Errors page for that time.) (3) Did the page reload or show an error message? (4) Was a required field empty? The replay shows exactly what happened.
Replay data grows with traffic volume. Use retention settings to automatically clean up old replay data. For high-traffic sites, consider a shorter replay retention period (e.g., 30 days) to manage storage.
If "Respect Do Not Track" is enabled in Settings, visitors with DNT headers are not recorded. You can also exclude specific pages from replay in the tracking settings.
The Events Log shows every event that OpsIQ has recorded: page views, clicks, chat starts, ticket creations, sales, and custom events. Use it to debug integrations, verify webhook delivery, and understand the sequence of actions.
Fired every time a tracked page loads. Includes URL, title, referrer, and timestamp.
Fired when a visitor opens the chat widget and sends their first message.
Fired for each message in a chat conversation (both visitor and AI/agent).
Fired when a new support ticket is created (from chat, email, widget, or admin).
Fired when a reply is added to a ticket.
Fired when a purchase is completed through a connected platform.
Fired when a tracked form is submitted (if form tracking is enabled).
Fired when a visitor is identified via an identity token.
Fired when a login attempt fails.
You can send custom events from your website or server to OpsIQ. Custom events are useful for tracking actions specific to your business: button clicks, feature usage, video plays, downloads, etc.
// Send a custom event from the browser
window.OpsIQ = window.OpsIQ || [];
window.OpsIQ.push(["track", "plan_selected", {
plan: "pro",
billing: "annual",
value: 99
}]);// Record a server-side activity for a contact (PHP)
$ch = curl_init("https://your-opsiq-domain.com/api/v1.php");
curl_setopt_array($ch, [
CURLOPT_POST => true,
CURLOPT_RETURNTRANSFER => true,
CURLOPT_HTTPHEADER => [
"Authorization: Bearer opq_your_key",
"Content-Type: application/json",
],
CURLOPT_POSTFIELDS => json_encode([
"action" => "crm.activities.record",
"site_key" => "site_abc123",
"email" => "[email protected]",
"event" => "subscription.upgraded",
"value" => 50,
"currency" => "USD",
]),
]);
$response = curl_exec($ch);
curl_close($ch);Filter by specific event names: pageview, chat.started, ticket.created, etc.
Show events from a specific time period.
Search event data by keyword (URL, email, event property values).
Events follow the retention policy set in Settings. Default: 90 days. High-value events (purchases, security events) can be configured with longer retention.
Yes. When creating an experiment, you choose a conversion event. Any built-in or custom event can be used as the conversion goal.
Experiments let you test two versions of a page or feature and measure which one performs better. OpsIQ tracks impressions, conversions, and statistical significance.
Give it a name, define your primary metric (conversion event), and set the variants (control and test).
OpsIQ randomly assigns visitors to variants and tracks their behavior. Do not change the experiment while it is running.
After enough data is collected, check confidence level. 95% confidence means you can be 95% sure the difference is real, not random chance.
Created but not yet started. You can still edit variants and metrics.
Actively assigning visitors to variants. Do not edit while running.
Reached statistical significance or was manually stopped. Results are final.
Results have been reviewed and the experiment is no longer needed in the active list.
Decide what you are measuring before starting. Choosing a metric after seeing results introduces bias.
Add your team's IP addresses to the exclusion list so staff visits do not skew results.
Do not stop an experiment after 100 visitors. You need hundreds or thousands of visitors per variant for reliable results.
If you change the headline AND the button color, you will not know which change caused the improvement.
After the experiment ends, note which variant won and why you chose to implement it.
It depends on the baseline conversion rate and the effect size you want to detect. As a rule of thumb: at least 500 visitors per variant for a 20% effect size. For smaller effects, you need thousands.
Yes, but be careful. If two experiments affect the same page, their results may interact. It is safer to run experiments on different pages or features simultaneously.
JavaScript Errors collects client-side errors from your website visitors. A broken button, a failed API call, or a missing resource shows up here. Use it to find and fix bugs that affect real customers.
The JavaScript error text (e.g., "TypeError: Cannot read properties of undefined").
Which page the error occurred on.
Which browser and device type saw the error. Some errors are browser-specific.
The code location where the error occurred. Click to expand the full stack trace.
When this error first appeared.
When this error most recently occurred.
How many times this error has been recorded.
How many different visitor sessions experienced this error.
Not all JavaScript errors need immediate attention. Prioritize by: (1) errors on critical pages (checkout, signup, login), (2) errors with high session counts, (3) errors that appeared recently (possible regression from a code change), (4) errors on your own code (not third-party scripts).
The JS Errors page shows "TypeError: Cannot read properties of undefined (reading submit)" on /checkout, affecting 45 sessions this week.
This means 45 real customers could not complete checkout. Steps: (1) Open the page in the same browser/device combination. (2) Open browser console and reproduce the error. (3) Check the stack trace to find the exact line of code. (4) Fix the bug, deploy, and monitor the error count going to zero.
Yes, OpsIQ captures all JavaScript errors on the page, including errors from analytics scripts, chat widgets, and ad scripts. Filter by URL in the stack trace to focus on your own code.
You cannot suppress specific errors in OpsIQ, but you can filter the list by page URL or error message. Fix or ignore errors from browser extensions (they often show as errors from unknown sources).
Sales tracking connects your store or billing platform to OpsIQ so you can see revenue alongside visitor behavior. Every purchase is attributed to a visitor journey so you know which pages, campaigns, and channels drive revenue.
Revenue from orders completed today. Refreshes with each new order event from your connected platform.
Revenue aggregated over the period. Refunds are subtracted from these totals.
Line graph showing daily revenue over the selected date range. Look for spikes (promotions, campaigns) and dips (issues, seasonal).
Which traffic sources lead to the most purchases. "Direct" buyers often bookmarked your site. "Google" buyers found you through search.
The latest orders with customer email, amount, currency, and items purchased.
Go to Connectors > Shopify. Create a custom app in your Shopify admin (Settings > Apps and sales channels > Develop apps). Grant scopes: read_orders, read_customers, read_products. Copy the Admin API access token. In OpsIQ, enter your store domain (e.g., your-store.myshopify.com) and the access token. Click Test Connection. If successful, enable webhook sync for real-time order notifications.
Go to Connectors > WooCommerce. In your WordPress admin, go to WooCommerce > Settings > Advanced > REST API. Create a key with Read/Write permission. Copy the consumer key and consumer secret. In OpsIQ, enter your site URL, consumer key, and consumer secret. Click Test Connection.
Go to Connectors > BigCommerce. In your BigCommerce admin, go to Advanced Settings > API Accounts. Create a V2/V3 API Account with read scope for Orders, Customers, and Products. Copy the access token, client ID, and API path. Enter them in OpsIQ and test.
Go to Connectors > Magento. In your Magento admin, go to System > Integrations. Create a new integration with API resource access to Sales, Customers, and Catalog. Activate the integration and copy the access token. Enter it in OpsIQ with your Magento base URL.
Go to Connectors > PrestaShop. In your PrestaShop admin, go to Advanced Parameters > Webservice. Enable the webservice and create a new key. Grant access to orders, customers, and products. Copy the key and enter it in OpsIQ with your store URL.
Go to Connectors > OpenCart. Generate API credentials from your OpenCart admin panel. Enter the API URL, username, and key in OpsIQ. The connector uses OpenCart's REST API for order and customer data.
Go to Connectors > osCommerce. Configure database access or API credentials as documented in the connector settings. osCommerce integration typically requires direct database access or a custom API module.
Go to Connectors > Stripe. Copy your Stripe restricted API key from the Stripe Dashboard (Developers > API keys). The key needs read access to Customers, Charges, Invoices, and Subscriptions. Enter it in OpsIQ. Configure webhook endpoint for real-time events.
Go to Connectors > Paystack. Copy your Paystack secret key from the Paystack Dashboard. Enter it in OpsIQ. Configure webhook URL for transaction notifications.
Use the REST API or webhooks to send sales events from any platform. Send order.completed events with customer email, amount, currency, and items. See the Events API section for the payload format.
Shows which channels bring paying customers. If Google brings 1000 visitors and 50 sales, that is a 5% conversion rate. If email brings 200 visitors and 30 sales, that is 15% — email is more effective per visitor.
Refunded orders are tracked separately. They reduce your net revenue totals. If refund rates are high from a specific source, the traffic might be low quality.
OpsIQ stores the original transaction currency and amount. Multi-currency businesses see revenue in each currency. Normalized reporting uses the configured exchange rate.
Your pricing page gets 100 visitors per day but zero conversions.
Investigate: (1) Use Session Replay on pricing page visitors. Watch how they interact with the pricing table. (2) Check Top Pages for the pricing page bounce rate and time-on-page. (3) Check Chat conversations from pricing page visitors — what questions do they ask? (4) Compare your pricing page to competitors. (5) Try an A/B experiment with different pricing layouts.
You sent an email campaign with UTM tags. How do you measure its revenue impact?
Go to Sales and filter by source. Search for the campaign UTM parameter. You will see: (1) How many visitors came from the email. (2) How many converted to sales. (3) Total revenue attributed to the campaign. (4) Compare to the same period before the campaign.
Check: (1) Is the platform connector connected and healthy? Go to Connectors and check the status. (2) Is webhook sync enabled? If using polling, is the cron job running? (3) Does the connector have the right API scopes? For Shopify, you need read_orders.
When an order comes in via the connector, OpsIQ matches the customer email to a known visitor. If the visitor was tracked before the purchase, the sale is attributed to their visitor journey, including the traffic source, landing page, and session history.
You can send sales events via the REST API or webhooks from any system. The event format is documented in the Events API section.
Leads are visitors or customers who show buying or support intent. OpsIQ automatically detects leads based on behavior: pricing page visits, return visits, chat interactions, checkout starts, and high engagement. Use the Leads page to prioritize follow-up and feed your CRM pipeline.
Visitors who view your pricing or plans page are actively considering a purchase.
Visitors who come back multiple times are engaged but not yet converted.
Visitors who start a chat are actively seeking help — either pre-purchase or post-purchase.
Visitors who reach the checkout page but do not complete the purchase.
Visitors who submit a contact form, demo request, or signup form.
Visitors who view many pages with long session duration are deeply researching.
Click any lead to see their full profile:
Name, email, company if known. Anonymous leads show IP and location.
All sessions from this lead with page flow, duration, and source.
Every page they visited, ordered by recency.
How they found you: organic, paid, social, referral, direct.
A 0-100 score based on behavior intensity, recency, and page intent.
Whether this lead has been added to the CRM as a contact or deal.
AI-generated suggestion on what to do next (e.g., "Visited pricing 3 times — consider proactive outreach").
Leads with a score above 80 get a "High Value" badge. These are your hottest prospects. Prioritize them for outreach, proactive chat, or CRM pipeline entry.
A lead viewed your pricing page 3 times in 2 days, started a chat asking about enterprise plans, but has not purchased.
This is a high-intent lead. Steps: (1) Check their lead detail for company and role information. (2) Review the chat transcript for specific needs and objections. (3) Create a deal in the CRM with the information gathered. (4) Assign a follow-up task for a sales team member. (5) If they return to the pricing page, the proactive chat rule can trigger a personalized message.
You have 50 anonymous leads with high scores but no email addresses.
Options: (1) Create a proactive rule that offers a discount or resource download in exchange for email on high-intent pages. (2) Use the chat widget to ask for email when anonymous visitors start a conversation. (3) Add an identity token to your login/signup pages so returning visitors are automatically identified.
The score considers: recency of visits (recent = higher), frequency of visits (more = higher), pages viewed (pricing/checkout = higher), session duration (longer = higher), chat interactions (engaged = higher), and return visits (multiple sessions = higher).
Yes. In CRM Configuration, you can add lead-scoring rules that boost the score based on specific keywords, events, sources, or page patterns.
OpsIQ ships a full AI-first CRM that captures customer truth, cleans data, scores leads, coaches deals, forecasts revenue, and acts — with website intelligence, support, tickets, products and payments as one live graph. Every AI action is sourced, explainable, reversible and approval-controlled.
Your daily home screen: revenue pulse, what AI found, what it did, approval queue, next moves, agent health, and ROI.
People and companies in your CRM. Search, filter, score, and manage profiles.
Visual deal board with drag-and-drop stages, win probability, and AI coaching.
Board-grade forecasting: committed, likely, best case, with confidence intervals and accuracy tracking.
Where every customer stands: active, at risk, expansion ready, renewal due, or churning.
Describe your CRM setup in plain English. OpsIQ generates a reversible plan and applies it on approval.
Self-check page: schema, agents, routes, AI actions, and live data counts.
AI proposals go to the approval queue. You approve, edit, or reject each one. Nothing happens without your say.
When AI suggests a change, it shows the current value and the proposed value. Human-entered data has provenance protection.
Every AI action is recorded in the journal. You can undo any change with one click.
Every AI recommendation shows its reasoning and the data sources it used. No black box.
Click AI CRM in the left navigation. The Command Center shows your current state.
Open CRM Health to verify the schema, agents, and routes are ready.
If a connector is active (Shopify, WHMCS, etc.), contacts are imported automatically. Otherwise, import via CSV or create manually.
Go to Pipeline and create your first pipeline with stages (e.g., Lead, Qualified, Proposal, Negotiation, Won, Lost).
Create a deal from a contact, set the amount and stage, and let the Deal Coach guide you.
No. The CRM is optional. OpsIQ works perfectly as a tracking + support + AI platform without the CRM. Enable it when you are ready to manage sales and customer relationships.
Yes. You can create contacts and deals manually. Connectors enrich the CRM with platform data (orders, subscriptions, tickets) but are not required.
The Trust Layer is the safety system that controls what the AI can do. Copilot mode (default): AI proposes, human approves. Autopilot mode (opt-in): AI acts automatically on low-risk, reversible operations.
The Command Center is your daily briefing. Open it every morning to see what happened, what needs attention, and what to do next.
Pipeline value, deals won/lost, new leads, active deals, and revenue trend — a snapshot of your current state.
AI-discovered insights: stalled deals, at-risk accounts, new opportunities, data quality issues, and market signals.
Actions the AI has taken (in Autopilot) or proposed (in Copilot): stage advances, score updates, contact enrichment, and task creation.
Queue of AI proposals waiting for your decision. Each shows the proposed action, reasoning, and evidence. Approve, edit, or reject.
AI-ranked list of the 10 most important things to do today: follow up on a hot deal, check an at-risk account, research a new lead, etc.
Health of AI agents: Capture, Data Steward, Deal Intelligence, Deal Coach, SDR, Retention. Green = running, amber = needs attention, red = error.
ROI tracking: deals influenced by AI, time saved, accuracy of predictions, and comparison to manual CRM work.
Check pipeline health and revenue trend. Are you on track this month?
Review each AI proposal. Approve safe actions, edit where needed, reject bad suggestions.
Read AI insights. Are any deals at risk? Are there new opportunities?
Start with the top priority. The AI has ranked these by urgency and potential impact.
You were away for a week. What happened while you were gone?
Open the Command Center and check: (1) "What OpsIQ did" shows all AI actions during your absence. (2) "Needs approval" shows proposals waiting for your review. (3) "What OpsIQ found" highlights urgent items. (4) Set the dashboard date range to your absence period to see the trend.
The CRM organizes your business relationships into four objects: Contacts (people), Companies (organizations), Deals (revenue opportunities), and Tasks (follow-up actions).
A person in your CRM. They might be a customer, a prospect, a partner, or a lead. Contacts have an email, name, company, tags, lifecycle stage, and hotness score.
Use the search bar to find by name, email, company, or tags. Use filters: lifecycle stage, lead score, source, country, last activity date, assigned owner.
Click a contact to see their full profile: personal info, company, deals, activities, tickets, chats, website visits, lead score, and AI insights.
A 0-100 score calculated from behavior signals: website visits, email engagement, chat interactions, purchase history, and recency. Higher = hotter.
Contacts are linked to companies by email domain or manual assignment. One company can have many contacts.
A revenue opportunity. Deals have a title, amount, pipeline, stage, win probability, close date, assigned owner, and associated contacts/company.
A percentage chance of winning the deal. Set manually or calculated by AI based on your historical win data. Drives the weighted forecast.
Green (healthy), amber (at risk), or red (stalled). Based on activity recency, stage duration, and engagement signals.
A follow-up action assigned to a person. Tasks have a title, due date, priority, type (call, email, meeting, other), and linked contact/deal.
The AI creates tasks automatically when it detects something needs attention: follow up on a stalled deal, respond to an at-risk account, research a new lead.
You want to see all high-score contacts in the UK who have not been contacted recently.
Go to Contacts. Filter by: Lead score > 80, Country = United Kingdom, Last activity > 7 days ago. Sort by lead score descending. These are your hottest neglected leads — prioritize outreach.
A contact just expressed interest in your enterprise plan during a chat.
Open the contact profile. Click "Create deal." Enter the deal title (e.g., "Acme Corp - Enterprise Plan"), amount, pipeline, and stage (e.g., "Qualified"). Set the close date and assign it to yourself. The Deal Coach will begin tracking this opportunity.
The Pipeline is a visual board where deals move through stages from left to right. Drag and drop cards to advance deals. Each card shows the deal amount, win probability, and health status.
The pipeline shows all active deals as cards arranged in columns (stages). Each column header shows the total value of deals in that stage.
Drag a card to the next stage. The win probability and counts update instantly. If the move fails (validation error), the card snaps back to its original position.
Green badge = healthy (recent activity, on track). Amber = at risk (slowing down, needs attention). Red = stalled (no activity for too long).
Click a card to open the deal detail panel with contacts, activities, Deal Coach brief, and quick actions.
AI proposals appear in the approval queue at the top of the pipeline page. Each proposal shows what the AI wants to do, why, and the evidence. You can:
Accept the proposal as-is. The action runs through the Trust Layer and is recorded in the journal.
Modify the proposal before approving. For example, change the suggested stage or adjust the amount.
Decline the proposal. The AI learns from rejections to make better suggestions in the future.
Win probability can be set manually or calculated by AI. When AI calculates it, it uses a model trained on your own closed deals (won and lost). If you don't have enough historical data, the AI uses the stage default probability. The probability drives the weighted forecast: a $10,000 deal at 30% probability contributes $3,000 to the "likely" forecast.
A customer signed the contract and paid.
Drag the deal card to the "Won" stage. OpsIQ will: (1) Mark the deal as closed-won. (2) Update the forecast. (3) Add a "Won" activity to the timeline. (4) If configured, trigger a webhook event (crm.deal.won). (5) Update the contact lifecycle stage to "Customer."
You want to find deals that have been in the same stage for too long.
Look for red (stalled) badges on the pipeline board. These deals have had no activity for longer than the configured rotting threshold (set in CRM Configuration > Pipelines). Click each stalled deal to see the Deal Coach's recommendation for reactivation.
Yes. Go to CRM Configuration > Pipelines to create additional pipelines. Common use cases: separate pipelines for new business vs. renewals, or for different products/services.
The deal is marked as closed-lost. You will be asked for a loss reason. The deal is removed from the active board but can be found in the closed deals filter. The contact remains in the CRM.
The Deal Coach analyzes each deal and gives you a brief with health status, risk signals, key people, a qualification checklist, meeting prep, and suggested next actions. Everything is sourced — the coach shows which ticket, chat, or activity each claim comes from.
Why the deal is healthy, at risk, or stalled. Specific signals: days in stage, last contact date, activity frequency, competitor mentions, objection patterns.
Categorized into: objections raised, competitor mentions, pricing discussions, support risk indicators, renewal/expansion signals. Each signal links to its source (chat transcript, ticket, email, activity).
Who is helping the deal (champion) and who is resisting (blocker). Based on interaction analysis and contact roles.
MEDDICC or BANT checklist (configurable) showing which qualification criteria are met and which are missing. Grade: A, B, C, D, or F.
Before a scheduled meeting, the coach generates a brief: agenda, key points to cover, questions to ask, risks to address, and materials to prepare.
Suggested next message or call script based on the current deal state and recent interactions.
Every claim the coach makes links to a source: a ticket number, chat transcript ID, activity date, or website visit.
Open any deal and scroll to the Deal Coach section. Read the health status and key signals.
Check the qualification checklist. If "Economic Buyer" is missing, plan to identify and engage the budget holder.
The coach suggests concrete next actions. Click to create a task, draft an email, or schedule a meeting — all routed through the approval queue.
You have a meeting with a prospect tomorrow about a $50,000 deal.
Open the deal and read the Meeting Prep section. It shows: (1) Key talking points based on recent interactions. (2) Objections they raised in chat last week. (3) Competitor they mentioned in a support ticket. (4) Questions to ask about their timeline. (5) Risk factors to address proactively. Use this brief to prepare a focused, evidence-based agenda.
The pipeline shows a red badge on a deal you thought was going well.
Click the deal and read the Deal Coach health explanation. It might say: "This deal has been in Proposal stage for 22 days (threshold: 14). Last contact was 11 days ago. The primary contact opened a support ticket about data migration concerns 5 days ago (ticket #T-892). Suggestion: address the migration concern and schedule a follow-up call." Now you know exactly what to fix.
Board-grade forecasting with a defensible range, deal-level change explanations, and accuracy tracking so you can prove and explain your numbers.
Deals with 80% or higher win probability. These are your most likely wins. This should be your minimum expected revenue.
All open deals weighted by their AI-calculated win probability. This is the most realistic prediction of what you will close.
All open deals at full value. This is the maximum possible revenue if everything goes perfectly (it rarely does).
An 80% probability band showing the realistic range. Example: "You will likely close between $45,000 and $72,000 this quarter."
The forecast includes week-over-week change attribution. When your forecast drops by $15,000, the report explains exactly which deals caused the change: "Deal A slipped to next quarter ($8,000), Deal B lost ($5,000), Deal C reduced in value ($2,000)."
Mean Absolute Percentage Error — how far off your forecasts have been historically. Lower is better. A MAPE of 15% means your forecasts are typically within 15% of actual results.
Brier score measuring how well win probabilities match actual outcomes. If deals at 70% probability actually win 70% of the time, calibration is good.
OpsIQ captures a snapshot of your forecast each week (via cron). Over time, you can see how your forecast accuracy improves.
How many deals move from each stage to the next. If 80% pass from Qualified to Proposal but only 20% go from Proposal to Negotiation, your proposals might need work.
Which lead sources produce the most wins. Invest more in high-win-rate sources.
How each sales team member is performing: deals won, pipeline value, average deal size, win rate.
Sandbagging (unrealistically low probabilities), overcommitting (unrealistically high), and committed-at-risk (high-probability deals showing stall signals).
Your quarterly forecast dropped from $180,000 to $155,000 since last week.
Open the Forecast page and check "What changed." The report shows: "Acme Corp deal ($15,000) moved to Lost — competitor chosen. Beta Inc deal ($8,000) close date pushed to next quarter. New deal with Gamma LLC ($3,000) added this week." The net change is -$20,000 + $3,000 = -$17,000, explaining the drop from $180K to $163K (with the remaining $8K explained by probability adjustments on other deals).
The Prospecting Agent helps you find and research potential customers. Describe your ideal customer, add target companies, and let the AI research, score, and draft first-touch outreach.
Go to CRM Configuration > Prospecting. Describe your ideal customer profile (ICP): target industries, company sizes, job titles, geographic focus, product-match terms, and minimum fit score. Example: "B2B SaaS companies with 50-500 employees in the US/UK, selling developer tools, with a VP of Sales or Head of Growth as the primary contact."
Enter company domains or names you want to research. You can also let the AI suggest companies based on your ICP.
The AI enriches each company with available data: size, industry, tech stack, recent news, and fit score with explanation.
For good-fit companies, the AI drafts a personalized first-touch message. Review and approve in the queue.
Each prospect gets a 0-100 fit score based on five factors:
How well the company matches your ideal customer profile (industry, size, location).
Is there evidence they need your product now? Job postings, tech changes, complaints about competitors.
Company size, funding stage, and revenue indicators suggest budget availability.
How relevant your product is to their business based on industry and technology.
Whether you can effectively serve this customer (timezone, language, complexity).
You want to evaluate whether acme.com is worth pursuing.
Add acme.com to the prospecting list. The AI returns: Fit score 78/100 (good fit). Reasons: B2B SaaS, 120 employees, US-based, recently hired a VP of Sales (urgency signal), uses competitor X (product match). Buying signals: visited your pricing page twice last month (from website tracking). Best contact: [email protected] (VP of Sales). Drafted opener references their recent hiring and pricing page visits.
Sequences automate multi-step follow-up: a mix of email, wait, task, call, and webhook steps that run automatically after you approve the initial send.
Define the steps: email, wait 3 days, email, wait 5 days, task (call if no reply), email. Each email step has a template you can customize.
Add contacts to the sequence. The AI may also propose enrollment from prospecting results.
Steps run automatically on schedule. If a contact replies at any point, the sequence pauses automatically.
See open rates, reply rates, conversion rates, and unsubscribes per step. Use this to optimize your sequence.
When a contact replies, the AI classifies the reply:
Interested, wants to talk. Sequence pauses, task created for follow-up.
Has concerns but not rejecting. Sequence pauses, task created with objection details.
Wants to stop receiving messages. Contact is suppressed permanently.
Interested but timing is wrong. Sequence pauses, snooze task created.
Not the right contact. Sequence stops, contact flagged for review.
A support question, not a sales response. Sequence pauses, ticket created.
Unsubscribe requests are honored immediately and the contact is suppressed from all future sequences.
Workspace-level daily limit (default: 1,000 emails). Prevents mass sends that trigger spam filters.
Maximum 1 message per contact per day. Prevents overwhelming individual people.
Steps that exceed the daily cap are deferred to the next day. They are never dropped or skipped.
The sequence pauses automatically. Support takes priority over outreach. The sequence can be manually resumed after the support issue is resolved.
Not directly within sequences, but you can create two sequences with different messaging and compare their results.
Lifecycle management helps you protect existing revenue. Each account gets a churn risk score, a health score, an expansion score, and a recommended success playbook.
Visual board showing where every account stands: Lead, Prospect, Opportunity, Customer, Expansion, Renewal, At Risk, Churned.
Accounts flagged with high churn risk. Each shows the top driver (support spike, inactivity, payment failure) and a recommended save play.
Accounts that show signals of readiness to buy more: high usage, feature requests, plan inquiries.
Accounts with upcoming renewal dates. Sorted by risk level and revenue value.
Churn risk is calculated from multiple signals fused together:
Deal health, inactivity, recent losses, declining engagement.
Open ticket count, ticket volume spikes, negative sentiment, escalations.
Login frequency, feature usage, page visits, API calls.
Failed payments, overdue invoices, downgrade requests, cancellation page visits.
For at-risk accounts: reach out proactively, address the concern, offer support, escalate if needed.
For upcoming renewals: confirm satisfaction, review usage, present value, offer incentive if appropriate.
For ready-to-grow accounts: present upgrade options, share success stories, demonstrate ROI.
For underutilizing accounts: offer training, share best practices, enable features they are not using.
A customer who normally logs in daily has not logged in for 2 weeks, and they opened 3 support tickets this month.
The Lifecycle board flags this account as "At Risk" with drivers: inactivity (14 days since last login) and support spike (3x normal ticket volume). Recommended play: proactive outreach call to understand what is happening. Is there a product issue? Are they evaluating competitors? Early intervention can save the account.
The Data Steward agent automatically finds and fixes data quality issues: duplicates, missing fields, inconsistencies, and stale records.
Multiple records for the same person (e.g., [email protected] and [email protected]). The steward proposes merging them.
Contacts without email, deals without amounts, companies without domains. Suggests filling from available sources.
Deals without associated contacts, contacts without companies. Proposes linking based on email domain and context.
A deal in "Won" stage with 20% probability, a contact marked "Customer" with no deals, a company with mismatched industry.
Deals that have been in the same stage for too long or contacts whose lifecycle stage does not match their activity.
Go to CRM and find the Data Steward section (or open it from CRM Health).
Green items are safe: clear duplicates, obvious missing data. Approve these in bulk.
Amber items need your judgment: possible duplicates with slight differences, proposed stage changes, company linking. Review each one.
The steward found "John Smith ([email protected])" and "J. Smith ([email protected])" — same email, different display names.
The steward proposes merging: keep the record with more activity, merge the other record's deals and activities. Review the proposal, adjust the surviving display name if needed, and approve. Both records' history is preserved in the merged contact.
Build My CRM lets you describe your CRM setup in plain English. OpsIQ generates a validated, reversible plan and applies it on approval. No coding, no manual configuration clicking.
Type a plain-English description of your CRM setup. Example: "Create a renewal pipeline with stages: Upcoming, Contacted, Negotiating, Renewed, Lost. Score leads higher when they view pricing twice."
OpsIQ shows a preview of what it will create: pipeline, stages, stage probabilities, scoring rules, lifecycle rules. Review each item.
Click Apply to execute the plan. Every change is recorded as a bundle you can undo in one click.
If something is not right, click Undo to revert the entire bundle. Your CRM returns to its previous state.
Create named pipelines with custom stages, default probabilities, and rotting thresholds.
Boost lead scores based on behavior: "Score higher when they view pricing," "Score lower when they visit careers page."
Automatically advance lifecycle stages: "Move to SQL when they request a demo," "Move to Customer when deal is won."
Add fields to contacts, deals, or companies: text, number, date, select, checkbox, URL, email.
Define the MEDDICC or BANT checklist the Deal Coach grades against.
You manage SaaS subscriptions and need a pipeline for renewals.
Type: "Create a renewal pipeline. Stages: 90 Days Out (10%), 60 Days Out (25%), Contacted (40%), Negotiating (60%), Renewed (100%), Churned (0%). Set the rotting threshold to 14 days." OpsIQ generates the plan, you review it, and click Apply. The pipeline appears on your Pipeline page immediately.
The builder only generates actions it knows how to validate. If you ask for something outside its capabilities (e.g., "Integrate with my custom API"), it will reject the request and explain what it can do instead.
Yes. Every Build My CRM action is captured as a reversible bundle. Go to CRM Health > Change History to see all bundles and undo any of them.
Manual CRM configuration for teams that prefer clicking over typing. All settings that Build My CRM can create are also available as traditional form fields.
Create, rename, reorder, and delete pipelines and stages. Set default win probability per stage. Set the rotting threshold (days without activity before a deal is marked "stalled").
Add fields to deals, contacts, or companies. Field types: text, number, date, select (dropdown with predefined options), checkbox, URL, email. Custom fields appear on the object profile and in filters.
Define the checklist the Deal Coach uses to evaluate deals. Choose MEDDICC, BANT, or a custom set of criteria. Each item can be marked required or optional.
Define automatic stage transitions: when a contact meets a condition (views pricing, submits a form, makes a purchase), their lifecycle stage advances automatically.
Boost or reduce lead scores based on behavior. Examples: "Viewed pricing page = +15 points," "Downloaded whitepaper = +10 points," "Visited careers page = -5 points." Minimum count can be set (e.g., "must view pricing at least 2 times").
Create named filter+sort combinations for contacts, deals, or companies. Share views with your team. Set a default view per object type.
You want contacts who view your pricing page twice or more to get a higher score.
Go to CRM Configuration > Lead Scoring Rules. Click Add Rule. Set: Event = "pageview", URL contains = "/pricing", Minimum count = 2, Score boost = +20. Save. The next time the scorer runs, contacts matching this rule will get +20 to their lead score.
CRM Health is a self-diagnostic page that verifies your CRM is set up correctly and running smoothly.
Database tables exist and have the correct columns. If a migration was missed, CRM Health will flag it.
All CRM agents (Capture, Data Steward, Deal Intelligence, etc.) are registered and healthy. Green = running, amber = needs attention, red = error.
CRM API routes and AJAX endpoints are registered and callable.
The CRM action catalog is registered in the Action Registry. If actions are missing, CRM Health self-heals by re-registering them.
Total contacts, companies, deals, tasks, activities, and events. Useful for verifying imports and ongoing data capture.
After first-time CRM setup, open CRM Health and verify all checks are green. If any are amber or red, click the item for a diagnostic explanation and suggested fix.
Check after initial setup, after major updates, and whenever CRM behavior seems wrong. CRM Health self-heals many issues (like missing action registrations) just by opening the page.
The CRM exposes a complete public REST API and outbound webhooks so it fits any stack without manual exports.
API key (Authorization: Bearer opq_...) with scope crm.read or crm.write. POST to /api/v1.php with action and site_key.
crm.contacts.list (search + filter), crm.contacts.detail (by ID), crm.contacts.upsert (create or update), crm.contacts.save.
crm.deals.list, crm.deals.detail, crm.deals.create, crm.deals.update, crm.deals.advance (all through the Trust Layer).
crm.activities.list (timeline for a contact/deal), crm.activities.record (log a new activity).
crm.events.list (durable change feed with since_id cursor), crm.events.catalog (list available event types).
POST /api/v1.php
Authorization: Bearer opq_live_xxx
Content-Type: application/json
{
"action": "crm.deals.create",
"site_key": "<workspace>",
"title": "Acme renewal",
"amount": 12000,
"pipeline_id": 1,
"stage_id": 2
}Fired when a new contact is added to the CRM.
Fired when a deal moves to a different stage.
Fired when a deal is closed as won.
Fired when a deal is closed as lost.
Fired when a new task is created (manual or AI-generated).
Fired when a prospect replies to a sequence email.
Fired when an AI agent creates a proposal for the approval queue.
All webhook events are HMAC-signed and retried on failure. Subscribe a URL in Settings > Webhooks, or poll crm.events.list with since_id for a durable change feed.
You want every new CRM contact to appear in a Google Sheet for your marketing team.
Set up a webhook subscriber for crm.contact.created. Point it to a Google Apps Script web app or a Zapier webhook URL. The payload includes the contact name, email, company, lead score, and source. Your script appends a row to the Google Sheet for each new contact.
Segments save reusable contact filters for CRM views, outreach, promotions and other audience-aware tools.
A live segment recalculates from current data whenever it is used. A pinned snapshot freezes the current member list, which is useful when an outreach audience must not change after approval.
Choose ALL when every condition must match or ANY when one condition is enough. Review the live count before saving. System segments can be visible but protected from ordinary editing.
Lead-scoring rules turn behaviour and profile signals into a consistent score that sales and automation can use.
The AI rule builder can turn a plain-English instruction into a structured rule. Every rule should still be reviewed before activation: confirm the activity or keyword, minimum count, rolling window and score change.
Positive and negative rules can coexist. Use Rescore workspace after material rule changes so existing contacts are evaluated under the new model. Scores can feed segments, routing, prospecting and workflow conditions.
The leaderboard compares representatives using quota attainment, won revenue, commission and open pipeline for the selected month.
Quota attainment is meaningful only when each representative has a current quota and deals have the correct owner, amount, currency and close status. Use the leaderboard for coaching and recognition, not as the only quality measure; pair it with deal health, activity and forecast accuracy.
Deal approvals require manager sign-off before high-value deals can be marked won.
Enable the policy and set Level 1 and Level 2 thresholds. A threshold of 0 disables that tier. When a non-manager tries to win a deal at or above the threshold, OpsIQ creates an approval request instead of silently changing the stage.
Managers and full administrators are not blocked by the policy. Every decision should retain the requester, approver, time and reason for auditability.
Lifecycle rules automatically move contacts between stages when their current data matches a saved condition.
Each rule has a source stage, destination stage, active flag, priority and segment-style filter. The scheduled lifecycle job evaluates higher priorities first and stops after the first match for a contact.
Rules should be mutually understandable. Avoid two active rules that can move the same contact in opposite directions during the same run. Test the filter against a segment before activating it.
The workspace activity stream combines customer and operational events so teams can investigate what happened without opening every module.
Filter by event type, date, contact or customer state (including paying customers). The feed is a read model of recorded activity; it does not replace the source ticket, conversation, deal or connector record. Open the linked record before taking a sensitive action.
Workflows respond to an event and run an ordered series of actions, including delayed steps resumed by cron.
Create a workflow by choosing the trigger, adding steps in order and enabling it. Delays are stored, not held by a browser request; the unified cron resumes due workflow runs.
Built-in starting templates include welcome after signup, post-purchase thank-you and demo confirmation. Clone a template, replace its recipients/content, then test with a controlled record before broad activation.
Design actions to be idempotent. A retry must not send duplicate money-moving requests or create duplicate records.
Tickets are your formal support channel. They have departments, priorities, SLA targets, attachments, internal notes, and full conversation threading. Tickets can arrive from the OpsIQ admin, ticket form/embed, email ingestion, chat escalation, connector, or API.
A customer sends a message via email, chat escalation, ticket embed, or the admin creates one. The ticket gets a unique ID (e.g., T-12345).
AI reads the subject and body, assigns priority (Low/Normal/High/Urgent), suggests a department, and detects sentiment. This happens automatically if triage is enabled.
An agent opens the ticket, reads the context (customer profile, platform data from connectors, previous tickets), and writes a reply. AI can draft the reply, and the agent edits before sending.
When the issue is resolved, the agent closes the ticket. The customer is notified. If the customer replies after closing, the ticket reopens automatically.
Move the ticket to a different department (e.g., from General to Billing). Useful when the initial routing was wrong.
Assign the ticket to a specific team member. The assigned agent sees the ticket in their "My tickets" view.
Write a note visible only to staff. Use for: investigation findings, refund reasons, engineering context, previous commitments. Internal notes never appear to customers.
Combine duplicate tickets from the same customer. The merged ticket subject gets a [MERGED] tag. All messages from both tickets are preserved in the surviving ticket.
Insert a pre-written reply template. Variables like {{customer_name}} and {{ticket_id}} are automatically replaced.
Let AI improve your draft reply: adjust tone, shorten/lengthen, fix grammar, or translate. You review the result before sending.
Set to Low, Normal, High, or Urgent. Priority affects SLA timers and sort order.
Set to Open, Pending, Waiting on Customer, Resolved, or Closed.
Subject: Double charge on my account Body: Hi, I just checked my bank statement and I see two charges of $49.99 from your company dated June 3. I only have one subscription. Can you please refund the duplicate charge? This is really frustrating.
Step 1: AI triage assigns Priority: High, Department: Billing, Sentiment: Frustrated. Step 2: Agent opens the ticket. The connector panel shows the customer's Stripe charges — confirmed two charges on June 3. Step 3: Agent adds an internal note: "Confirmed duplicate charge in Stripe. Transaction IDs: ch_xxx and ch_yyy." Step 4: Agent drafts a reply: "Hi [Name], I can see the duplicate charge and I'm processing a refund for the second transaction right now. You should see $49.99 back in your account within 3-5 business days. I'm sorry for the inconvenience." Step 5: Agent triggers the refund action through the Stripe connector (with confirmation). Step 6: Agent sends the reply and sets status to Resolved.
Create departments in Settings > Tickets > Departments. Common departments: General, Billing, Technical, Sales, Abuse, Onboarding. Each department can have different auto-reply rules, AI knowledge, and team assignments.
Create reusable reply templates in Settings > Tickets > Canned Responses. Good canned responses include:
Hi {{customer_name}},
Thanks for reaching out. I checked the details and here is what I found:
{{response_details}}
If you need anything else, just reply to this ticket and I will be happy to help.
Best regards,
{{agent_name}}
{{department_signature}}When a customer sends the same question multiple times (via email and chat, or two separate emails), you can merge the duplicates. Go to the ticket, click Merge, select the other ticket to merge. The surviving ticket gets all messages from both tickets. The merged ticket's subject gets a [MERGED] tag to indicate it was combined.
Non-urgent questions, feature requests, general feedback. Target first response: 24 hours.
Standard support questions. Target first response: 8 hours.
Impacting the customer's business. Target first response: 4 hours.
Service down, security incident, data loss, billing error. Target first response: 1 hour.
SLA timers start when the ticket is created and pause when the status is "Waiting on Customer." If the SLA target is missed, the ticket is highlighted and optionally escalated.
Never. Internal notes are strictly staff-only. They do not appear in the customer ticket portal, email notifications, or any customer-facing surface.
The ticket automatically reopens and returns to the active queue. The team is notified of the new reply.
Yes. Use the Lock Thread option to prevent both customer and AI from adding new replies. Useful for resolved disputes or closed investigations.
The control room for how your support operation behaves: SLA policies with escalation ladders, business-hours calendars, an enforced status flow, CSAT settings and the enterprise feature switches. Open it from the Tickets page (the Service Studio button in the operations bar).
Every new ticket gets a first-reply and a resolution deadline from the most specific active policy: department + priority beats department, beats priority, beats the default. Deadlines pause automatically while a ticket sits in Pending or On hold.
Instead of one breach action, a policy can define up to six steps: for example warn the assignee at 75% of the window, alert managers at 100%, reassign to a named agent at 150%. Each step fires exactly once per ticket.
With business hours on, SLA clocks only burn during working time, so a ticket opened Friday evening is not breached by Saturday. Named calendars (timezone, days, hours, holidays) can be attached to individual policies for teams in different regions.
A from/to matrix restricting which status changes admins may make, with a Zendesk-style preset. Automation, inbound email and customer-originated ticket replies are always exempt, so a strict matrix can never jam the pipeline. Optionally require an approved sign-off before closing.
Agents request sign-off on a ticket (refunds over a threshold, policy exceptions). A named approver or any Full Administrator decides with a note; requesters can never approve their own request. Pending approvals surface as a banner on the inbox.
Link tickets as related, duplicate, parent/child or blocking; split replies out to a new linked ticket (the reverse of merge, without emailing the customer); follow tickets for in-app notifications, and optionally CC follower email addresses on outbound replies.
With the real-time inbox enabled, new tickets and replies appear for agents within seconds through the existing notification stream. On a ticket, agents see who else is viewing and get a warning banner when a teammate is already writing a reply. A one-click Claim button assigns the ticket to yourself.
Routing rules can match on AI triage signals: priority, category, sentiment and escalation risk, plus typed custom-field values from your portal request form. Rules using these conditions stay dormant until the feature is enabled.
The matrix only governs ADMIN actions. Customer-originated replies, email replies and automation are deliberately exempt so a strict policy can never lock customers out or break the pipeline.
No. Policies saved with a single breach action keep working unchanged. A ladder replaces the single action only on policies where you define steps.
Auto-reply lets the AI respond to new tickets automatically. Escalation moves tickets to a different department when the AI detects specialized needs. Both features work across native tickets and email-created tickets.
A new ticket is created from any source (email, chat, portal, API).
OpsIQ waits for the configured delay (default: 2 minutes). This gives human agents a chance to respond first.
If a human agent replies during the delay, auto-reply is cancelled. If not, the AI generates a reply.
The AI reads the ticket, checks knowledge base, connector context, and customer history, then writes a reply. The reply is sent (or queued as a draft for review, depending on settings).
Set which departments can receive escalations (e.g., Billing can be escalated from General, Technical can be escalated from General).
Before auto-replying, the AI classifies the ticket intent: billing, technical, refund, abuse, etc.
If the ticket intent matches an escalation department, the ticket is moved. An internal note explains why.
The reply is generated using the destination department's knowledge and AI training.
Master switch. Turn on to enable AI auto-reply for tickets.
Which departments can receive auto-replies. Start with low-risk departments (General, Sales) before enabling billing.
How long to wait before the AI replies (in seconds). Default: 120 seconds (2 minutes). Set to 0 for instant replies.
Optionally set different delays for business hours vs. after hours.
If a human agent replies first, cancel the auto-reply. Default: on.
Enable automatic department routing based on ticket intent.
Which departments tickets can be escalated to.
What status to set when escalating (e.g., Open, Pending).
Whether to add an internal note explaining the escalation reason.
Tickets created from email (via mailbox polling or MX-piped delivery) follow the same auto-reply and escalation rules. The AI reply is sent as an email response to the customer. Email drafts go to the drafts queue for review if configured.
Only if you enable auto-reply for the Billing department. We recommend starting with auto-reply off for billing and enabling it only after you have trained the AI with comprehensive billing knowledge and tested thoroughly.
If a customer replies saying the answer was wrong, a human agent should intervene. Review AI History to understand why the AI gave the wrong answer, then update the knowledge base or AI training to prevent it from happening again.
Yes. Set the delivery mode to "Draft" instead of "Auto-send." Auto-reply drafts will appear in the Drafts queue for human review before sending.
The ticket portal embed lets you add a ticket management interface to any website. Customers can view their tickets, create new ones, and reply — all without entering the OpsIQ admin.
Renders the ticket portal inside a div on your page. The portal takes the full width of its container.
Shows a floating button that opens the ticket portal in a panel. Similar to a chat widget.
Requires a signed identity token. Customers see only their own tickets.
<script src="https://your-opsiq-domain.com/widget.php" data-site-key="site_abc123" data-mode="tickets" data-identity-token="SERVER_GENERATED_TOKEN" async></script>
For authenticated mode, your server must generate a signed identity token. This token tells OpsIQ who the customer is so they see only their own tickets.
<?php
// Generate identity token (PHP)
$site_secret = "opsiq_whsec_your_secret";
$payload = json_encode([
"site_key" => "site_abc123",
"customer" => [
"id" => "cust_1001",
"email" => "[email protected]",
"name" => "Ada Lovelace",
],
"iat" => time(),
"exp" => time() + 900, // 15 minutes
]);
$token = base64_encode($payload) . "." . hash_hmac("sha256", $payload, $site_secret);
// Pass $token to the frontend// WHMCS hook — auto-generate identity token
add_hook("ClientAreaPage", 1, function($vars) {
if (!isset($vars["loggedinuser"])) return;
$payload = json_encode([
"site_key" => "site_abc123",
"customer" => [
"id" => $vars["loggedinuser"]["id"],
"email" => $vars["loggedinuser"]["email"],
"name" => $vars["loggedinuser"]["firstname"]." ".$vars["loggedinuser"]["lastname"],
],
"iat" => time(),
"exp" => time() + 900,
]);
$token = base64_encode($payload).".".hash_hmac("sha256", $payload, "opsiq_whsec_xxx");
// Inject into page
});// Node.js / Express — generate identity token
const crypto = require("crypto");
function generateOpsIQToken(customer) {
const payload = JSON.stringify({
site_key: "site_abc123",
customer: {
id: customer.id,
email: customer.email,
name: customer.name,
},
iat: Math.floor(Date.now() / 1000),
exp: Math.floor(Date.now() / 1000) + 900,
});
const sig = crypto.createHmac("sha256", "opsiq_whsec_xxx")
.update(payload).digest("hex");
return Buffer.from(payload).toString("base64") + "." + sig;
}Searchable, status-filtered list. An "Awaiting you" highlight flags tickets the team has answered.
Full conversation view with rich text, attachments, and timestamps.
Customer draft replies are auto-saved so they do not lose work if they navigate away.
The widget polls for new replies every ~12 seconds.
Customers can resolve or reopen their own tickets with one click.
The Chat Inbox is where your team manages live customer conversations. It shows active chats, assigned conversations, AI-handled chats, and archived history.
Conversations currently in progress. Green dot = customer is typing or recently active.
Conversations where the AI is responding. You can take over at any time by clicking "Join."
Conversations assigned to you specifically.
Conversations where the customer is waiting for a response.
Completed conversations. Searchable by keyword, date, and customer.
When you take over a conversation from the AI, follow these three rules:
Introduce yourself by name. "Hi, this is Sarah from the support team."
Show you read the conversation. "I can see you asked about your order status."
Tell the customer what happens next. "I am looking into this now and will have an answer for you in a few minutes."
Use the AI toggle at the top of each conversation. When AI is on, the AI generates responses. When off, only human agents can reply. The AI remembers the conversation context even when turned off and back on.
Configure business hours in Settings. Outside business hours, the AI can: (1) Continue responding to customers, (2) Show an "away" message and create a ticket for follow-up, or (3) Disable chat completely and show the ticket form. Choose the behavior that fits your support model.
Do not match the customer's emotional intensity. Acknowledge their frustration without being defensive.
Before responding, gather the relevant information: order details, account status, previous tickets.
If the situation involves refunds, legal threats, abuse, or security, escalate to a senior team member rather than trying to handle it alone.
Yes. Use internal notes within a conversation to communicate with other agents. These notes are never visible to the customer.
The conversation moves to the Archive. If the customer sends a new message later, a new conversation is created.
These examples show how a well-trained Customer AI should handle common scenarios. Use them to test your AI training and identify gaps.
Customer: Hi, I placed an order 3 days ago and haven't received any shipping notification yet. Order #ORD-5523.
The AI should: (1) Look up order #ORD-5523 through the connector. (2) Report the current status and tracking info if available. (3) If the order is delayed, acknowledge the delay and provide an estimated timeline. (4) Do NOT invent a shipping status if the connector does not have one.
Customer: How much does your Pro plan cost? Do you offer annual billing?
The AI should: (1) Answer from the knowledge base with current pricing. (2) Explain the difference between monthly and annual billing. (3) If the answer is not in the knowledge base, say so and offer to connect them with sales. (4) Do NOT invent pricing that is not in the knowledge base.
Customer: This is ridiculous! Your product does not work and I want my money back NOW.
The AI should: (1) Acknowledge the frustration calmly. (2) Ask for the order number or account email. (3) Explain the refund process. (4) Create a support ticket or escalate to a human agent. (5) Do NOT promise a refund — that requires human approval.
Customer: I was charged twice this month. My bank shows two charges of $29.99.
The AI should: (1) Acknowledge the concern. (2) Collect the relevant details (email, dates of charges). (3) Explain that billing issues require human review. (4) Create a Billing department ticket with the details. (5) Set priority to High.
Customer: Bonjour, je voudrais savoir si vous livrez en France?
The AI should: (1) Respond in the customer's language if that language is supported. (2) Answer the question about delivery to France from the knowledge base. (3) If the language is not supported, respond in the default language and mention which languages are available.
Customer: Hello, is anyone there? I need help with my account.
The AI should: (1) If configured for after-hours AI, respond and assist. (2) If after-hours creates tickets, say: "Our team is currently offline. I have created a support ticket for you and someone will follow up during business hours." (3) Provide the ticket number.
Customer: How does your product compare to [Competitor X]?
The AI should: (1) Describe your product's strengths factually. (2) Do NOT badmouth the competitor. (3) If there is a comparison page in the knowledge base, link to it. (4) If the customer asks about features you do not have, be honest.
Customer: I run a small online store with about 100 orders per month. Which plan is right for me?
The AI should: (1) Ask clarifying questions if needed (e.g., what features matter most). (2) Recommend a plan based on the knowledge base. (3) Explain why that plan fits their needs. (4) Mention that they can start with a smaller plan and upgrade later.
Customer: I saw online that you give 50% discounts. Can I have one?
The AI should: (1) Politely explain current promotions if any exist. (2) Do NOT create or promise discounts that are not in the knowledge base. (3) If the customer persists, offer to connect them with the sales team.
Customer: How do I export my data? I can't find the export button.
The AI should: (1) Provide step-by-step instructions from the knowledge base. (2) Include the exact page path and button location. (3) If the feature requires a specific plan, mention that. (4) If the feature does not exist, be honest and suggest alternatives.
OpsIQ can connect to email mailboxes to automatically create tickets from incoming email and send replies as email. It also provides AI-drafted replies for human review.
IMAP (most providers), Gmail API (Google Workspace), or Microsoft Graph (Outlook 365).
Host, port, encryption (SSL/TLS), username, password or app password, folder (usually INBOX).
Map the mailbox to a ticket department. Emails to [email protected] create tickets in the "Support" department.
Go to Google Account > Security > 2-Step Verification and enable it.
Go to Google Account > Security > App Passwords. Select "Mail" and "Other (OpsIQ)". Copy the 16-character password.
Go to Email > Mailboxes > Add. Set host: imap.gmail.com, port: 993, encryption: SSL, username: [email protected], password: the app password.
Click Test Connection. If successful, OpsIQ will start polling for new emails at the configured interval.
In Outlook settings > Mail > Sync email, ensure IMAP is enabled.
If MFA is enabled, create an app password. Otherwise, use your regular password.
Host: outlook.office365.com, port: 993, encryption: SSL, username: [email protected].
In cPanel > Email Accounts, click your email account for IMAP settings.
Host: usually mail.yourdomain.com or your server hostname, port: 993 (SSL) or 143 (STARTTLS), username: full email address.
When auto-reply is set to "Draft" mode, the AI generates reply drafts that appear in the Drafts queue. You can:
Read the draft and check for accuracy.
Modify the text, tone, or content before sending.
Approve the draft and send it as an email reply.
Delete the draft if it is not appropriate.
If confidence is high enough and auto-send is enabled, drafts are sent automatically after a configurable delay.
Check: (1) Is the mailbox connected and healthy? Look for green status in Email > Mailboxes. (2) Is the cron job running? Email polling requires cron. (3) Is the email being filtered or moved to a folder other than INBOX?
Yes. Each mailbox can be mapped to a different department. [email protected] goes to Support, [email protected] goes to Billing.
CSAT measures ticket and chat satisfaction, surfaces unhappy responses and compares AI and human outcomes.
Ticket surveys are scheduled after closure (normally 24 hours later). A positive response is a score of 4 or 5. Use the detractor queue for scores of 1 or 2 and read the ticket/conversation before contacting the customer.
Chat feedback is reported separately for AI-resolved and human-handled conversations. Low sample sizes should not be used to rank agents or models.
The Email Platform page lets an authorised operator send a one-time message to all active agents, selected agents or a department and review send history.
This page is for internal operational messages, not customer marketing. Choose the smallest correct audience, preview names and addresses, then send. Delivery uses the configured outbound email path and records the result.
Configure how OpsIQ sends email, turns inbound messages into tickets and proves the sending domain is ready.
Full administrators can choose the preferred email connector; OpsIQ falls back to configured SMTP/mail delivery when appropriate. Set a recognisable From name/address, send a test and check the 24-hour success/failure metrics.
An IMAP mailbox needs a name, email address, target department, host, port, security mode, username, password, folder, polling state and certificate-validation choice. Use Poll now to verify ingestion before waiting for cron.
On the same server, a mail pipe can pass the raw RFC822 message directly. A remote mail server can POST the raw message to the protected inbound endpoint using its token. Both paths enter the same deduplication, threading and ticket-routing pipeline.
The SPF/DKIM/DMARC check is advisory evidence for the configured From domain. A green check does not guarantee inbox placement, but a failure should be fixed before sending production messages.
Support Portal Studio designs and publishes the customer-facing support experience without requiring external image assets.
Use the template as a starting point, then configure theme, layout, navigation, knowledge search, ticket form, assistant, footer, brand, ticket behaviour, custom CSS and general settings. Preview desktop, tablet and mobile before publishing.
The knowledge-gap queue collects customer questions the AI could not answer confidently so the team can improve coverage.
Open a gap to read the original question and nearby conversation, then choose the right fix: update an existing article, create a new knowledge article, add a business fact, correct connector data or change escalation instructions. Mark it resolved only after an evaluation or real test returns an acceptable answer.
The status administration page communicates service health and incidents using components, current states and updates.
Use component states consistently. When an incident affects customers, publish an initial notice, add time-stamped updates as facts change, and close with a concise resolution summary. Do not use the status page as an internal debug log; sensitive diagnostics belong in OpsIQ and server logs.
AI Configuration is where you choose your AI provider, model, and cost controls. OpsIQ supports multiple providers: Claude (Anthropic), GPT-4o (OpenAI), Gemini (Google), and Grok (xAI). You can use your own API key (BYOK) or Managed AI if available on your plan.
Strong reasoning, careful with facts, good at following complex instructions. Models: claude-sonnet-4-20250514, claude-3.5-sonnet. Best for: support conversations, technical questions, detailed analysis.
Fast, creative, good general knowledge. Models: gpt-4o, gpt-4o-mini. Best for: quick responses, creative writing, broad knowledge questions.
Good at factual questions, multi-language support. Models: gemini-2.0-flash, gemini-1.5-pro. Best for: multi-language support, factual lookups.
Fast, concise responses. Models: grok-3-mini-fast. Best for: quick answers, casual tone.
You provide your own API key from the provider. You pay the provider directly. Full control over model selection and usage.
AI is provided as part of your OpsIQ plan. No API key needed. Usage is metered against your plan's AI credit balance.
In the OpsIQ admin, navigate to Settings > AI Configuration.
Choose your AI provider from the dropdown.
If using BYOK, enter your API key. For Managed AI, this is handled automatically.
Select the specific model to use. Smaller models are faster and cheaper. Larger models give better quality answers.
Click "Test connection" to verify the API key and model are working. A green checkmark means success.
Click Save to activate your AI configuration.
Set a maximum monthly spend on AI tokens. When the budget is reached, AI features gracefully degrade (shorter responses, no auto-reply) rather than stopping entirely.
Maximum tokens per single conversation. Prevents runaway conversations from consuming excessive credits.
When your AI credit balance drops below the threshold (default: 50,000 tokens), a warning appears in the admin. Set this in AI Configuration.
AI configuration is per-workspace. If you manage multiple sites, each workspace can use a different provider, model, and budget. The workspace inherits global defaults unless overridden. To configure per workspace: switch to the workspace, then go to AI Configuration and save settings there.
Go to console.anthropic.com > API Keys > Create Key. Copy the key starting with "sk-ant-".
Go to platform.openai.com > API Keys > Create new secret key. Copy the key starting with "sk-".
Go to aistudio.google.com > Get API Key. Create a key for your project. Copy the key.
Go to console.x.ai > API Keys. Create and copy your key.
After entering your API key, click "Test connection." You should see a green checkmark and a sample response from the AI.
If the test fails: (1) Check that the API key is correct (no extra spaces). (2) Check that your account has billing set up with the provider. (3) Check that the model name is valid. (4) If using Managed AI, verify your license is active.
There is no single best provider. Claude is excellent for support conversations. GPT-4o is fast and versatile. Gemini handles multiple languages well. Try each one and see which produces the best results for your specific use case.
Yes. Change the provider and API key at any time. Existing conversation history is preserved. The AI will use the new provider for all future conversations.
The AI degrades gracefully: responses become shorter, auto-reply pauses, and the admin shows a low-balance warning. Core tracking, tickets, and CRM continue working without AI.
A set of opt-in controls that make your customer chat either smarter (it reasons across a customer's account like a human agent) or cheaper (it sends the AI less, or answers in fewer steps). Every control is off by default and independent — turning one on never changes any other behaviour. Most live under Settings > Client Chat > Behaviour; a few are on Set up your AI and Train your AI.
A normal chatbot answers one question from one lookup. With Agent mode on, the assistant can make several read-only account checks in a single turn and cross-reference them — e.g. "do I have open tickets, and is my balance enough for my next renewal?" is answered in one reply with the maths done. It only engages for verified customers on account or refund questions; everything else runs the normal single-answer path.
Up to a few read-only lookups per turn, cross-referenced before answering. Turn on when you have a connected platform (WHMCS, Shopify, Stripe, Zendesk, etc.). Off = today's single-lookup behaviour, unchanged.
How hard the model reasons on complex turns (low/medium/high). Start on Low; most support questions are handled well there.
Pins a stronger model just for complex agent turns, so everyday chat stays on a light, cheap model and only the hard turns pay for depth. Blank = your normal customer-chat model.
A tiny planning step decides which lookups to pull and whether the knowledge base is needed, then answers in ONE main call instead of several steps. A sub-toggle under Agent mode — it cannot be enabled unless Agent mode is on. Same answer quality, fewer calls on multi-part questions.
Keeps a short running summary of older messages (names, ticket/invoice numbers, amounts, decisions) so the assistant stays consistent across very long chats without resending the whole transcript. Best for support-heavy, long threads; pairs with Agent mode.
Reads small typos ("my open ticketts", "unpaid invocies") as the intended words so the right information is found instead of missed. Works with every connected platform. Off by default; useful for mobile-heavy chat.
Only attaches the customer's account snapshot when the question is actually about their account (balance, invoices, tickets, services), skipping it on general questions. Cuts cost with no quality loss — the assistant can still fetch the account live when a turn needs it.
"Hi", "thanks" and "ok" don't trigger a knowledge-base search — those replies never use it. Any message with real content still searches as normal.
Caps how long a single reply can be. Reply length is the most expensive part of a call, so a sensible cap (around 500 for support) is the single most effective chat cost control. Only lowers your global limit, never raises it.
How much the model thinks before answering (minimal to high). Keep customer chat on minimal/low for speed; reserve higher settings for analysis and the admin copilot.
Every AI surface — live chat, ticket drafts, summaries, analysis, admin copilot, promotions, Site Intelligence and more — can run on its own model. Light where speed matters, strong where quality matters. Blank on a surface = provider default for that job.
The admin copilot can keep a bounded working memory and load only the tools a question needs (lean tools) instead of every tool every turn.
Your instructions are sent on every turn, so their length is a recurring cost. The Compress button rewrites a slot to say the same thing in fewer words. It refuses the rewrite if it would drop any {variable} or rule, and reports what it protected — so it never silently changes meaning.
Shows what OpsIQ already knows about your business per slot, so you can delete anything you were repeating in your own instructions.
No. Every control is off by default and independent. Turning one on only adds its behaviour; turning it off returns the chat to exactly how it worked before.
Run everyday chat on a light model and pin a stronger Agent model. Then, if you have a connected platform, enable Agent mode, then Plan ahead, then Long-conversation memory. Then lean context, skip-KB-for-greetings, and a reply cap. Finally compress your instructions. Watch a few real chats after each change.
The Admin AI chat is a working engineer, not just a question box. It can research fixes on the web, keep a library of your operating procedures, patrol your servers on a schedule, and move files. Most of this is driven by asking for it in chat; the one thing you configure up front is which web search provider it uses, on the Set up your AI page.
The agent searches the web for docs, error messages and current information while it works. Out of the box this uses DuckDuckGo, which needs no key and no setup.
A saved library of your procedures ("how we do X here"). The agent writes one down after nailing a tricky job and follows it next time, so the know-how survives the end of the chat.
Scheduled read-only patrols. "Every morning, check disk space and the error logs and tell me." Findings arrive on your admin bell.
Ask for a log file and get a private download link, or have one file copied from one connected server to another. Both capped at 5 MB.
What the agent web search uses. The default is DuckDuckGo: keyless, free, works on every install. Brave Search, Tavily and Serper.dev (Google results) usually return better results but each needs an API key from that provider. If a keyed provider fails for any reason (bad key, quota used up, network trouble), that search falls back to DuckDuckGo automatically, so search never goes dark.
The key for your chosen provider, only needed for Brave, Tavily or Serper. Once saved it is masked and never shown on the page again; paste a new one to replace it. A keyed provider becomes active only when both the provider and its key are saved.
When the agent has just worked through a job you will want done the same way again, say "save that as a runbook called nightly-cleanup". From then on it sees the runbook by name in every session, and when you ask for the job by name (or describe it) it pulls up the steps and follows them. You can ask it to list, update or delete runbooks any time.
A mission is a recurring job the agent runs on its own: hourly, daily at a set hour, or weekly on a set day. Missions are read-only by design. They can read servers, data and the web, but they can never change anything; they investigate and report. When a mission finishes, the findings land as a notification on your admin bell. Say "run the morning patrol now" to trigger one on demand, or ask the agent to list, pause or remove missions.
Ask for a file ("get me today's error log from the web server") and the agent stages it and replies with a private download link. The link only works for signed-in admins and expires after about a day. Files are capped at 5 MB; for anything bigger, ask the agent to compress or trim it on the server first.
The agent can also copy a single file between two connected servers, again up to 5 MB. The write on the destination is treated like any other change it makes: in ask mode you get an approval card first, the existing file is backed up, and the copy is reversible. For safety, the file content is re-read from the source at the moment the copy is applied, not taken from the chat.
AI Training is where you teach OpsIQ how to behave. You write prompts that define personality, boundaries, knowledge focus, and escalation rules. Think of it as writing a training manual for a new support agent.
Controls how the AI talks to your customers in the chat widget. This is the most important prompt to get right.
Controls how the AI assists your admin team in the operator dashboard. This handles internal queries and action execution.
Your customer AI prompt should cover these areas:
Who is the AI? Give it a name, role, and personality. Example: "You are Luna, a friendly and knowledgeable customer support assistant for Acme Store."
How should it communicate? Example: "Be professional but warm. Use simple language. Avoid jargon. Keep responses concise (2-3 paragraphs max)."
What should it know and not know? Example: "You know about our products, pricing, shipping, and return policy. You do NOT know about competitor products or internal company decisions."
When should it hand off to a human? Example: "Escalate to a human agent when: the customer asks for a refund, mentions legal action, reports a security issue, or asks the same question three times."
Hard boundaries. Example: "Never promise refunds. Never share internal pricing formulas. Never make up product features that do not exist. Never give medical, legal, or financial advice."
You are Aria, a friendly customer support assistant for TechGear. PERSONALITY: - Warm and professional - Concise (2-3 paragraphs max per response) - Use simple language, no jargon - If unsure, say so honestly KNOWLEDGE: - Our products: laptops, accessories, software subscriptions - Pricing: as listed in the knowledge base - Shipping: free over $50, 3-5 business days standard, 1-2 days express ($12) - Returns: 30-day money-back guarantee, must be in original condition ESCALATION (hand off to human): - Refund requests over $100 - Account security concerns - Legal threats or complaints - Customer explicitly asks for a human NEVER DO: - Promise specific delivery dates (say "estimated") - Share discount codes not in the knowledge base - Make up product specifications - Give financial or legal advice
You are Atlas, an AI assistant for CloudHost web hosting. TONE: Technical but accessible. Assume the customer has basic web knowledge. KNOWLEDGE: - Shared hosting, VPS, dedicated servers - cPanel, DNS, SSL, email configuration - WordPress, Joomla, Drupal installation - Common error codes and troubleshooting ESCALATION: - Server outages affecting multiple customers - Data loss or security breach reports - Billing disputes over $50 - Complex migration requests BOUNDARIES: - Do NOT access customer servers or databases - Do NOT change DNS records or passwords - Always recommend the customer makes changes themselves - For code issues, suggest steps but clarify you cannot debug their application
Start with identity, tone, and key boundaries. Do not try to cover everything on day one.
Read conversations where the AI gave wrong or suboptimal answers.
Add specific rules for recurring issues. Example: if the AI keeps promising refunds, add "Never promise a refund — say you will escalate to the billing team."
If the AI cannot answer a question, add the answer to the Knowledge Base rather than trying to fit everything in the prompt.
The prompt defines personality, boundaries, and rules. The knowledge base provides factual information (product details, pricing, policies). The AI reads both, but the prompt shapes behavior while the KB provides answers.
The prompt is per-workspace. If you need different AI behavior for different departments, use the knowledge base to provide department-specific information and mention the department context in your prompt.
AI History shows every conversation the AI has had with customers and admins. Use it to audit AI performance, find wrong answers, and identify training gaps.
When the conversation happened.
Name and email if known, otherwise IP and country.
Chat widget, ticket auto-reply, email draft, or admin AI.
Full transcript of the conversation: customer messages and AI responses.
How many tokens the conversation consumed. High token usage may indicate the AI is giving overly verbose answers or the conversation went in circles.
Any actions the AI executed during the conversation (looked up order, created ticket, etc.).
If CSAT is enabled, the customer's satisfaction rating for this conversation.
Search for specific topics to see how the AI handles them. Example: search for "refund" to check if the AI is following your refund policy.
Show only low-rated conversations to focus on dissatisfied customers.
Show only ticket auto-replies to audit automated responses separately from chat.
High-token conversations may indicate problems: the AI rambling, going in circles, or failing to answer the question.
Read the full transcript. What went wrong? Was it a knowledge gap, a tone issue, or a wrong answer?
These might indicate confusion. Was the AI repeating itself? Did it fail to understand the question?
For knowledge gaps, add articles to the Knowledge Base. For behavior issues, update the AI prompt. For recurring wrong answers, add explicit correction rules.
You search AI History for "installation" and find 15 conversations where the AI said "I do not have specific installation instructions for this product."
This means your Knowledge Base is missing installation guides. Create knowledge base articles for each product's installation steps. After adding them, test by asking the AI installation questions.
AI Insights automatically analyzes conversation patterns and surfaces actionable findings: common questions, satisfaction trends, knowledge gaps, and escalation patterns.
The most frequently asked questions. If "How do I reset my password?" is #1, your knowledge base needs a prominent password reset guide.
CSAT scores over time. Is AI satisfaction improving or declining? Sudden drops may indicate a knowledge base change that broke something.
What percentage of conversations does the AI resolve without human intervention? A rate above 70% is good. Below 50% means the AI needs more training.
Why conversations get handed off to humans. Common reasons: customer requested human, AI could not answer, sensitive topic, billing issue.
Topics where the AI frequently says "I don't know" or gives low-confidence answers.
How many messages per conversation. Very long conversations (10+ messages) might indicate the AI is not understanding the question.
AI Insights shows that 23% of escalations are "Customer asked about warranty" and the AI could not answer.
This is a clear knowledge gap. Steps: (1) Write a comprehensive warranty article in the Knowledge Base covering all warranty terms, claim process, and exclusions. (2) Update the AI prompt to reference warranty information. (3) Monitor AI Insights next week to see if warranty escalations decrease.
The Admin AI (Ask OpsIQ) helps your team manage OpsIQ from within the admin dashboard. You can ask it to look up data, explain features, summarize tickets, and execute platform actions.
Ask: "Show me all tickets from [email protected]" or "What is the status of order #ORD-5523?"
Ask: "Summarize the last 10 tickets" or "What are the top support issues this week?"
Ask: "How do I set up auto-reply?" or "What does the bounce rate card mean?"
Ask: "Create a ticket for [email protected] about billing" or "Close ticket T-12345"
Ask: "How many visitors did we get yesterday?" or "What is our revenue this month?"
Admin asks: "Find customer [email protected] and show me their recent activity"
The AI looks up the customer across all connected data sources: visitor sessions, tickets, chat conversations, orders, and CRM contacts. It returns a summary of recent activity with links to each item.
Admin asks: "What happened today?"
The AI summarizes: "Today you had 245 visitors (up 12%), 8 new tickets (3 resolved by AI), 2 new orders totaling $198, and 1 chat escalation about a shipping delay."
Admin asks: "Change ticket T-5523 priority to Urgent"
The AI confirms: "I will change ticket T-5523 priority from Normal to Urgent. This will affect the SLA timer. Confirm?" You confirm, and the action is executed.
This section covers common customer-facing AI scenarios with expected behavior and training tips.
Customer: "Where is my order #12345?"
Expected AI behavior: (1) Look up the order through the platform connector. (2) Report current status (processing, shipped, delivered). (3) Include tracking number if available. (4) If the order is delayed, acknowledge and provide estimated timeline. (5) Never invent tracking information.
Customer: "I forgot my password and cannot log in"
Expected AI behavior: (1) Provide the password reset steps from the knowledge base. (2) Include the reset link or page path. (3) Mention that the reset email may take a few minutes. (4) Suggest checking spam folder. (5) If the customer still cannot reset, escalate to a human.
Customer: "Can you add dark mode to the mobile app?"
Expected AI behavior: (1) Thank them for the suggestion. (2) Explain how feature requests are handled (logged, reviewed, prioritized). (3) Do NOT promise the feature will be built. (4) If a similar feature exists, mention it.
Customer: "I was charged but my service is not working"
Expected AI behavior: (1) Acknowledge the frustration. (2) Check the service status if possible. (3) Escalate to the billing department with all details. (4) Do NOT promise a refund. (5) Set ticket priority to High.
Customer: "What is the weather like in London today?"
Expected AI behavior: (1) Politely redirect. "I am your [Company] support assistant and can help with questions about our products and services. For weather information, try a weather service." (2) Ask if they have a product question.
Open your website in a private/incognito browser and chat with the AI as a customer. Try common questions, edge cases, and adversarial scenarios. Check that the AI follows your prompt rules and uses knowledge base content correctly.
The Knowledge Base is a library of articles that the AI uses to answer questions. When a customer asks something, the AI searches the knowledge base for relevant articles and uses them to generate an accurate, grounded response.
"What is your return policy?"
OpsIQ finds the "Return Policy" article by matching keywords and semantic meaning.
The AI reads the article and writes a natural-language answer based on the article content.
The response is grounded in your actual policy, not the AI's general training data.
Write a separate article for each topic: return policy, shipping rates, account setup, password reset. Do not combine unrelated topics.
Start with the question customers ask, then provide the answer. This helps the AI match questions to articles.
"Returns must be initiated within 30 days of delivery" is better than "We have a return policy."
What about international returns? What about digital products? What about items bought on sale? Cover the exceptions.
Review articles quarterly. Delete outdated information. Update pricing, policies, and procedures when they change.
Shipping costs, delivery times, tracking, international shipping, and handling delays.
Return window, conditions, process, refund timeline, and exceptions.
Password reset, email change, account deletion, and data export.
Plan descriptions, billing cycle, payment methods, invoices, and currency.
How to use each product or feature. Step-by-step instructions with screenshots.
Common problems and solutions. Error messages and their fixes.
Business hours, support channels, response times, and escalation paths.
Privacy policy summary, terms of service summary, GDPR data requests.
Each connector can contribute knowledge to the AI. When a connector is enabled, its built-in knowledge articles (product catalog, common questions, platform-specific troubleshooting) are automatically available to the AI. You can toggle connector knowledge on or off per connector in the Connector settings.
The knowledge base has a configurable token ceiling (default: 32,000 tokens) that controls how much knowledge context is sent to the AI per conversation. If your knowledge base is very large, the AI selects the most relevant articles within this budget. Increase the ceiling if the AI is not finding relevant articles; decrease it if you want to reduce token costs.
Start with 5-10 articles covering your most common questions. Add more as you identify gaps through AI History and AI Insights. Most businesses need 20-50 articles for comprehensive coverage.
Currently, articles are created manually in OpsIQ. You can copy-paste content from existing FAQs, help centers, or documents.
Plain text works best. The AI understands natural language better than structured HTML or markdown. Write as you would explain to a customer.
Evaluation sets turn critical customer questions into repeatable tests for answer quality, citations, escalation and action safety.
Full administrators can create evaluation sets with a prompt, expected source labels, required citations, human-confirmation rules, escalation expectations and forbidden actions. Run a set directly, dry-run it, or replay selected history.
Watchtower schedules daily or weekly checks through cron and runs a bounded number of cases (up to eight per scheduled pass). Review failures alongside provider/model, prompt version, knowledge version, actions proposed and cost before changing production behaviour.
Languages manages translated customer-facing text and the behaviour used when OpsIQ follows a customer’s language.
OpsIQ translates operator-entered customer-facing text. Set the source language correctly before creating translations, add target languages, then review drafts before publishing. “Follow customer language” controls response language; translating the widget interface controls its buttons, labels and system text.
Translation scope can be draft-only or live. Do not rely on machine translation for legal, refund, security or safety text without human review.
Your Help Center is a self-serve help site built from the articles you publish. Customers can read and search it on its own public page, and right inside the chat widget, so they can find answers on their own before they ever open a conversation. Everything in it comes from your knowledge base, so the same content that grounds your AI also powers your public help.
A clean, searchable help site at <code>/help.php</code>. Customers browse by category, search, and open any article. You can give it your own name and style.
A help link in the chat widget so customers can search your articles without leaving the chat. If they still need a person, they are one tap away from starting a conversation.
Go to <code>Config, then Help Center</code> in your admin and switch on <strong>Enable public Help Center</strong>.
Set the Help Center name, which shows in the header and the browser tab, and pick a visual style. You can turn the top navigation off for a clean page with no menu.
Nothing shows until you publish articles to it. The next part covers how.
Articles live in your knowledge base. An article appears in the Help Center once it is published and marked public. There are a few ways to do this:
Open the article, mark it published, and turn on the option to show it in the Help Center.
Use the <strong>Publish All</strong> button to push your whole published set to the Help Center in one click, instead of doing them one at a time.
Already have FAQs? Use <strong>Publish all to Help</strong> to turn them into help articles so they sit alongside everything else.
Moving from another help desk? The Import KB button on the Articles tab brings an existing knowledge base in for you. You have three ways to point it at your content:
Give it the address of your current help center. If it recognizes the platform (for example Zendesk) it reads the exact category tree from that platform. Otherwise it crawls the site and rebuilds the structure from the pages, importing only the real articles and skipping login, pricing, and contact pages.
If a platform is already connected to this workspace and it holds a knowledge base, it shows up in a dropdown under <strong>Import from a connected app</strong>. Select it and import, no URL or password needed. The app is only listed once it is connected and set up.
No live site to point at? Upload a spreadsheet with <code>title</code>, <code>category</code>, and <code>body</code> columns. Put <code>Parent / Child</code> in the category column to build nested categories.
Whichever way you choose, you get a preview first: the articles it found, grouped by category, each with a checkbox so you can drop anything you do not want. Nothing is saved until you click Import. Afterwards you can Undo the whole import in one click, or Re-sync a URL or connected-app import later to pull in new and changed articles.
A tidy help center is easy to browse and easier to search. The Organize with AI button does the tidying for you. In one click it reads your published articles, sorts them into sensible categories, and adds tags. It creates a category when it needs one, reuses a category that already fits, and skips anything you have already organized, so you can run it again any time you add new articles. Turn on Let AI organise into subcategories in Help Center settings and it can also file articles under nested Parent / Child categories, which is handy right after importing a knowledge base that had sub-sections.
You can show the Help Center right inside the chat widget so people search answers before they message you. In your chat widget settings, choose where the help link appears:
A small help icon at the top of the widget.
A text link under the message box, for example "Browse help center". You can change the wording.
Show it in the header and at the bottom for the most visibility.
The help link has its own color control under the widget Appearance settings. Leave it on None to match your launcher color, or pick a Solid color or a Gradient. The whole help experience follows your widget theme, so it always looks like part of your brand.
Out of the box your help center already has a clean, shareable address of its own (for example /help/your-help-center). You can go one step further and serve it under your own domain, so customers and search engines see it as part of your brand — for example yourdomain.com/help or help.yourdomain.com. Everything keeps working on your domain: links, images, search and the article ratings.
In <strong>Config, then Help Center</strong>, open the <strong>Embed</strong> tab and find "Show your help center on your own web address". Pick your setup — Cloudflare, Nginx or Apache — copy the ready-made block, and paste it into your site. Each option has a <strong>Setup guide</strong> button next to it with the exact steps.
Prefer <code>help.yourdomain.com</code>? The same panel has a subdomain option. On Cloudflare a subdomain is the most reliable route, and it gives you a clean address with nothing after it.
Beyond single-article votes, the Help Center has its own analytics. Switch it on in Help Center settings and it quietly records how people use your help — the most-viewed articles, what they search for, and searches that came back empty (a ready-made to-write list) — working out visitor country and device on its own, so it does not need the chat widget installed to report. Every number is shown for both places your help appears, the public page and the in-widget help, as well as combined, so you can see how much each one is actually used.
Each section of the Help Center — the categories on the home page, the popular articles, the article list on a category page and more — has its own simple rule. For each one you can show or hide it, and show all items, a set number, the latest, or only the ones you feature with a star. When a rule caps a list, customers still get a "Show all" link, so nothing is truly out of reach.
You also decide how the helpful votes look: hidden, a friendly ratio like "12 of 14 found this helpful", or plain thumbs counts. A minimum-votes setting stops a brand-new article showing an awkward "0 out of 0".
Every article shows a Was this helpful? prompt. Each yes or no is counted per article, so you can see at a glance which articles answer questions well and which ones need a rewrite. Repeated votes from the same visitor are ignored, so the counts stay honest.
You can change the headings and labels on the Help Center to match your voice. Leave any field blank and it falls back to a sensible default, so you only edit the wording you actually want to change.
No. It is built from the same knowledge base articles the AI already uses. Publish an article once and it can serve both.
Yes. Only articles you publish and mark public appear in the Help Center. Anything unpublished stays out of it.
Yes. The widget help link sits a tap away from starting a conversation, so self-serve never traps anyone.
Site Intelligence is OpsIQ's built-in SEO and search-visibility suite. It crawls your site like a search engine, tracks where you rank, researches keywords, watches backlinks and competitors, scans your local map presence, checks page speed, and turns it all into customer-ready reports — with an AI analyst that explains what to fix and can draft the fixes. It is a premium feature; when enabled it appears as Site Intelligence in the sidebar. Every domain's data stays separate — crawls, rankings and reports never mix between domains.
A crawl is OpsIQ visiting your pages the way a search engine would, to find technical issues: broken links, missing titles, thin content, slow pages, indexing problems.
On the Overview, click to start a crawl. OpsIQ queues it and works through your pages in the background; a status indicator shows progress.
When it finishes, open the issues and the page list. Each issue is grouped by severity so you tackle the biggest wins first.
The AI analyst produces a plain-language brief of what is wrong and why it matters — and can generate suggested fixes you review before applying.
Scans your pages for technical SEO issues and lists every page with its problems. Re-run any time, or schedule it.
Add the keywords you care about; OpsIQ checks where your site ranks and tracks movement over time.
Discover new keyword ideas to target, with the AI keyword coach to help prioritise.
See who links to you, refresh the list, and disavow links you do not want associated with your site.
Add competitor domains and compare your visibility against theirs, with an AI competitor analysis.
Scan local-listing visibility across a geographic grid to see where you show up on the map, plus Google Business Profile tools.
Run page-speed checks so you can fix slow pages that hurt rankings and conversions.
Generate content briefs and drafts aimed at the keywords you are targeting.
Build printable, customer-ready reports of everything above.
Two Google connectors turn estimates into real numbers:
Pull the exact queries, clicks and impressions Google reports for your site, map your property, and surface ranking opportunities straight from Google's own data.
Bring your traffic and engagement data alongside your SEO data, so reports show visibility and behaviour together.
You do not have to run everything by hand. Scheduled refreshes re-run crawls, rank checks, keyword checks, backlink refreshes, local-grid scans and report generation on a cadence (daily or weekly depending on plan and volume). These run through the OpsIQ cron — see the Cron and automation section. Generated reports export as printable, customer-ready documents: agencies send each client a regular SEO update, and because every domain's data is separate, each client only ever sees their own.
Crawls, rank tracking, keyword research and backlinks work without touching your site. For traffic and exact search data, connect Google Search Console and GA4.
Yes. Add each domain separately; their crawls, rankings and reports stay fully isolated from one another.
Weekly is plenty for most sites; crawl on demand after a big content or structure change. Set a schedule so you never forget.
Auto-Implement is the part of Site Intelligence that actually applies the fixes the crawl and AI analyst recommend, instead of only advising you. The agent works strictly from what Site Intelligence already found — it can never invent a change — and walks each fix through the same safe loop: a dry-run preview, your per-step approval, the live write, an automatic snapshot, and a re-crawl that proves the open-issue count went down. It is a Managed-AI-only feature; it appears as the Auto-Implement (rocket) action inside the Site Intelligence section, and it is off until you turn it on.
The agent only ever applies fixes that Site Intelligence (or the AI analyst) already recommended for that domain. Across the supported platforms that covers:
sitemap.xml, robots.txt, and redirect, canonical and header rules in .htaccess (Apache / LiteSpeed) or the nginx config.
Collapses a multi-hop redirect chain down to a single hop straight to the final URL, inside the same managed .htaccess block, so speed and link value stop leaking through the extra hops.
Missing or weak page titles, meta descriptions, canonical tags and structured-data (schema) blocks.
Yoast, Rank Math and SEOPress titles & meta, and image alt text, applied through WP-CLI on the SSH path or the REST API on the hosted path.
Adds JSON-LD (schema) to pages that are missing it. On the SSH path a small managed helper (a must-use plugin) prints the markup so it works with or without an SEO plugin; on the hosted path your SEO plugin carries it.
Adds one relevant internal link inside existing WordPress content to rescue orphan pages and connect related posts. One link per page, never inside a heading, an existing link or a shortcode, and always approved by you first.
Drafts full replacement content for thin pages (turned on by a crawl option, see below). Every draft is reviewed and approved by you before anything publishes, and the old content is kept so it can be restored.
Turns on W3 Total Cache page caching when the crawl finds speed problems and that plugin is installed, so pages serve faster.
Product and collection SEO titles & descriptions, page and blog-article SEO, plus 301 redirects, through the Shopify Admin API.
Per-page SEO title and meta description through the Webflow Data API v2.
The Site Intelligence Configure panel for a domain (its crawl settings) has two options that decide how much the agent lines up for you ahead of time. Both are off by default and both use AI credit, so you switch them on only when you want that work done for you.
Every crawl also writes the concrete fix content (titles, meta descriptions, alt text, schema, redirects) for what it found, so the fixes are ready to review and apply the moment the crawl finishes instead of being generated later on demand. Uses AI credit on each crawl.
The agent may draft full replacement content for thin pages. Nothing is ever published on its own: each draft appears as a change you review side by side and approve or reject, the previous content is kept so it is reversible, and it is limited to a few pages per pass. Uses AI credit per draft.
For sites on your own server, the agent connects over SSH/SFTP using credentials you store, edits the server config and page files directly, and runs WP-CLI for WordPress. Best when you have shell access and want full control of files.
For platforms that do not allow SSH, the agent works entirely through an official API connector — WordPress (REST + Application Passwords), Webflow (Data API v2) and Shopify (Admin API). Nothing is installed on the host.
The agent reaches each platform through a connector:
Shopify and WordPress connectors ship with OpsIQ — just enable them under Connectors and add the store / site credentials.
Webflow, Wix and Squarespace install from Connectors → Marketplace, then connect the same way.
Open Auto-Implement (rocket) from the Site Intelligence section. It lists the issues Site Intelligence already found that it can fix on this domain.
For every fix the agent shows a dry-run diff — exactly what it will write — before anything touches the live site.
You approve (or skip) each change individually. Nothing is applied without your explicit go-ahead on that step.
On approval the agent makes the live change over SSH or the hosted API and saves a snapshot first, so any change can be rolled back.
As it applies each change the agent immediately re-checks that exact page and shows a per-step “fixed live” verdict, so you watch each fix land on the live page as it happens. At the end it re-crawls the whole site and confirms the open-issue count dropped.
The agent can only do what Site Intelligence recommended for that domain — it cannot make arbitrary changes to your site or server.
Every write is previewed as a diff and applied only after you approve that specific step.
A snapshot is saved before each write, so any change can be reverted.
Each changed page is re-checked the instant it is written and gets a “fixed live” verdict, so a fix is confirmed against the live page right away, not only in the closing re-crawl.
Internal links and thin-page content rewrites are higher-impact, so they are always shown to you for approval and are never applied automatically, whatever mode you run.
Auto-Implement runs on Managed AI; it is blocked when a workspace is using its own (BYOK) keys.
AI tokens bill normally as you use them, plus one flat per-use fee per Auto-Implement session — both drawn from the same Managed-AI credit.
No. Every change is previewed as a dry-run diff and applied only after you approve that exact step. A snapshot is saved before each write so you can roll back.
Not on WordPress, Shopify or Webflow — those work through their official APIs with no shell access. SSH is only one of the two paths, used when you want the agent to edit server config and files directly on your own server.
After applying the approved fixes the agent re-crawls the site and shows the open-issue count before and after, so you can confirm the fixes reduced the issues.
No. AI content rewrites are off until you turn them on in the crawl options, they are limited to a few thin pages per pass, and every draft is shown to you for side-by-side approval before anything publishes. The previous content is kept, so a rewrite can always be reverted.
No. Auto-Implement is Managed-AI only and is blocked under BYOK. It bills AI tokens normally plus one flat per-use fee per session, from the same Managed-AI credit.
The Promotion Studio turns your website into a conversion engine. Design pop-ups, banners, slide-ins, top/bottom bars, full-screen takeovers, inline blocks and even country-based redirects — then show the right message to the right visitor at the right moment. Everything is delivered through the OpsIQ widget already on your site, so there is nothing extra to install. You design and save campaigns any time; they only show once you Publish them and the studio is switched on for your workspace.
One promotion — a pop-up, banner, bar, slide-in, full-screen, inline block or geo-redirect. Each has a goal: sales, lead, sign-up, chat, survey, announcement.
You build the look in the visual editor. Every save creates a new version, so you can always roll back — nothing is ever lost.
Who sees it: country, page, device, traffic source, returning vs new, schedule, and — with a connector — your logged-in customers.
When it appears: instantly, on delay, on scroll, on exit-intent, when idle, on a click, after N pages, or from chat / ticket activity.
Live impressions, click-through, conversions, a drop-off funnel, country / page / device splits, A/B winners and an AI insight.
Open Promotion Studio and click New campaign, or pick a ready-made design from the Templates gallery. A draft opens in the designer.
Drag in blocks — headings, text, buttons, images, badges, stats, countdowns, forms, coupons, videos, ratings, spin-wheels. Style colours, gradients and 3D effects with the sliders. Preview Desktop / Tablet / Mobile, and use Free move to place anything anywhere.
Type tokens like {first_name}, {country} or {company} into any text. At show-time they become the visitor's real values.
Choose who sees it (country, page, device, audience) and when it fires (delay, scroll, exit-intent). Leave a rule blank to mean everyone.
Click Check for a quick pre-launch scan, then Publish. It goes live within seconds and starts collecting analytics.
Heading, paragraph, list, quote, image, icon, badge, stat, divider and spacer — the building blocks of any message.
Buttons with real actions: open a link, start a chat, create a ticket, copy a coupon, or jump to a page.
Countdown timers, forms (email / phone / text), coupon reveals and video embeds.
Spin-to-win wheels, scratch cards, NPS / star ratings and multi-step funnels.
Columns, progress bars, social share, social proof tickers, yes/no two-step, quizzes, and 3D effects — tilt, glow, parallax, glassmorphism.
Type any of these in headings, paragraphs, buttons, badges, stats, lists or quotes. They are replaced with the visitor's live details the moment the promotion shows (always cleaned for safety):
From the visit: {country} {city} {device} {page_path} {utm_source} {referrer}
From your CRM: {first_name} {last_name} {email} {company} {segment} {lifecycle_stage}
From a connector: {balance_due} {unpaid_invoices} (logged-in customers only)Show only in (or hide from) chosen countries, cities or timezones via GeoIP. Great for region-specific offers or a geo-redirect.
Limit to certain pages using equals / starts-with / contains / regex, and exclude pages too.
Desktop, tablet or mobile — e.g. a slim bar on mobile, a modal on desktop.
New, returning, known (in your CRM) or anonymous; CRM segments, lifecycle stage, lead score or browser language.
By UTM (utm_source, utm_campaign) or referring site — e.g. only visitors from a specific ad.
Start / end dates, time-of-day windows, and how often a person can see it (per session, per day, after they close / click / convert).
Show to a percentage of visitors and run variants head-to-head; the same person always sees the same variant.
With a connector enabled, target by live account state — e.g. customers with unpaid invoices, or active subscribers.
Pick how the promotion appears: Immediately, after a delay, on scroll depth, on exit-intent, when idle, on a click of an element, after a number of pages viewed, on form-abandon, or from chat / ticket activity. Developers can fire one from their own code with window.OpsIQPromo.fire('event').
This is what makes OpsIQ promotions special: connect a platform like WHMCS (or any connector that supports audiences) and the studio can target visitors by their live account state.
Under Connectors, enable one that exposes audience facts (WHMCS does out of the box). Customers are recognised securely when logged into that platform — OpsIQ never trusts a typed-in email.
In the campaign's Targeting > Audience, require logged-in and add rules like unpaid invoices >= 1, subscription is past-due, or account credit > 0.
Use tokens such as {first_name} or {balance_due}. A "You have an overdue invoice" reminder writes itself.
A Product block pulls real items — name, price, sale price, image, link — straight from WHMCS, WooCommerce, Shopify or your OpsIQ billing. Tick "Keep fresh" and the card updates its price from the platform every hour while published.
A coupon block can create the real discount code in your store the moment you publish, so the code a visitor reveals always works at checkout.
Real orders from your platforms are matched back to campaigns hourly — by coupon code used or by the lead's email — so the revenue you see is from actual sales, not estimates.
The social-proof ticker can lead with genuine recent orders ("Ada ordered Starter Plan · 2h ago") — first name and item only, never emails or amounts.
Split traffic across design variants; let the studio promote the statistical winner automatically, or hand traffic to the bandit optimiser which shifts visitors toward what converts as evidence comes in.
Reserve a slice of visitors who see nothing, so you can prove the campaign itself causes the lift.
Different content per audience on the SAME campaign — mobile visitors, big carts, returning customers — each segment gets its own version.
Variants can also test WHEN to appear: immediate vs delayed vs exit-intent, measured head-to-head.
Target by what someone did with another campaign — "saw campaign A but didn't convert" — to build multi-step promotion sequences.
A small pill visitors can reopen after closing the campaign, so a closed pop-up is never gone for good.
In the designer, toggle the heatmap to paint real visitor clicks over every block of your design.
Multi-step campaigns show exactly where people drop off, screen by screen.
Put up to four campaigns side by side — impressions, clicks, conversions — best value highlighted.
Every Monday a digest lands in your notifications; the 📖 Weekly story button writes a plain-language summary with next steps whenever you want one.
Chat with the design copilot to edit the canvas ("make the headline urgent, add a countdown"), get campaign ideas from your own traffic data, translate a campaign per country, or auto-fix a low design score.
One command reads your website and proposes your real brand colours and fonts; save and every picker leads with them.
Steps blocks get true per-screen editing — tabs for each screen, add or reorder screens, and the preview follows along.
Name important versions ("Launch candidate"), restore any of them, and see a warning when a teammate is editing the same campaign.
See desktop, tablet and mobile side by side in one view; toggle pixel rulers for precise placement.
Every campaign can also be a standalone hosted page for bios and QR codes, an email drip sequence for captured leads, or a one-click web-push draft.
Publish freeze windows (change-freeze periods), an emergency pause-all switch, approvals, audit trail and version rollback.
No. Promotions are delivered by the same OpsIQ widget you already installed. Design, publish, done.
No — set frequency caps (per session / per day, or stop after someone closes, clicks or converts). Sensible defaults are applied.
Publishing, pausing, archiving and rolling back are limited to workspace owners and full administrators. Any team member can create and design drafts.
Promotion Designer is a responsive block editor for building, reviewing, versioning and publishing a campaign experience.
Use blocks to compose the experience, preview each supported viewport and insert approved personalisation variables. Review comments and AI-media provenance before launch. Saving creates versions so a previous design can be inspected or restored.
Templates provide a searchable, categorised starting gallery for Promotion Studio campaigns.
Search or filter by category, load additional results, preview the complete experience, then choose Use template. The new campaign remains a draft and can be changed in Promotion Designer before publication.
Promotion analytics connects impressions and clicks to conversions, audience context, connector action outcomes and campaign learning.
Read impressions, click-through rate and conversions together. Break results down by country, page, trigger and device; export captured leads when necessary. Connector action outcomes show whether the promotion’s promised follow-up actually ran. AI insight and learning summaries should be treated as hypotheses to validate, not automatic proof.
Announcements publish a message in the widget notification area with optional targeting, schedule and call to action.
Choose all visitors, signed-in visitors or new visitors; optionally narrow the audience with a contact segment and page rule. Set active-from/active-to dates and an active flag. Premium style presets and colours change presentation but do not replace clear content.
Push campaigns send browser notifications to subscribed visitors after they have left the site.
Configure Web Push keys first under the Web Push settings. A campaign can include title, body, icon, hero image, badge, CTA and up to two action buttons; a persistent notification remains until the visitor interacts.
Target by audience, optional contact segment and page rule. Save a draft, schedule it or send immediately. Use the subscription and click metrics to judge reach. HTTPS and a correctly hosted service worker are required on embedded third-party sites; see Web Push on third-party sites.
The Cookie Consent studio designs the cookie banner your visitors see, defines what counts as essential for your business, applies the right behaviour for each visitor's region, and — crucially — actually enforces consent so analytics and marketing tags do not run until the visitor agrees. The banner is delivered through the OpsIQ widget already on your site. It is a premium design studio; the banner shows only once you switch consent on for your workspace.
Pick from a large library of banner designs and a full colour system — transparent, solid or gradient — to match your brand.
Fine-tune wording, button labels, policy links, banner position, and how long a choice is remembered.
Define your own cookie categories and the wording visitors see. You decide what is essential — essential is always on; everything else is theirs to allow or refuse.
Scan your site and let OpsIQ AI auto-detect every cookie and tracker, categorise it, and write a plain-language purpose. This powers the preferences table and auto-blocking.
Choose how consent behaves by law (below), and turn on group-domain sharing, Google Consent Mode v2, and re-consent intervals.
See live consent stats from real visitor choices — accept rates, by category — scoped to your workspace.
EU/UK visitors get opt-in (nothing non-essential runs until they accept); US visitors get opt-out with a "Do Not Sell or Share" option. OpsIQ decides per visitor by location. Recommended.
Strict GDPR for all visitors — nothing non-essential until they accept.
Tags run unless the visitor opts out.
Browser Global Privacy Control (GPC) and Do-Not-Track (DNT) signals are honoured whichever mode you pick.
Consent here is not just a banner — it controls what runs:
<!-- This tag stays dormant until the visitor grants the "marketing" category -->
<script type="text/plain" data-opsiq-consent="marketing"
src="https://example.com/ads-pixel.js"></script>Turn on group-domain sharing so a visitor who accepts on one of your sites is not asked again on the others in this workspace group — their choice carries across for the number of days you set. Turn it off for strict per-domain consent. Cookie Consent pairs with the Privacy & Compliance page, which keeps the consent ledger (proof of every grant and revoke).
Yes. Per-category enforcement stops the matching capture at the source — it is not a cosmetic banner.
No for the banner — it ships through the widget. Only auto-blocking your own third-party tags needs a one-time markup change on those tags (shown above).
On the Privacy & Compliance page: a searchable consent ledger of every grant and revoke by contact, channel, action and date.
Proactive Messages reach out automatically the moment a visitor's behaviour matches a rule — before they ask. You build each as a rule with three parts: what should happen, when it should fire, and how often. Everything is delivered through the OpsIQ widget already on your site.
Show a chat bubble or on-page banner, start a Guided Tour, send a web push or email, assign to a department, or fire a webhook. Tokens like {{visitor_name}} and {{page_title}} become real values.
Narrow who triggers it: page/URL, time on page, scroll depth, lead score, returning vs new, idle. With no conditions, the rule fires for every visitor — add at least one to target it. Conditions combine with AND.
Set a cooldown so the same person is not pestered — e.g. once per session or once per day. Each rule has an on/off status.
Only if you add no conditions. Add at least one condition (page, lead score, returning…) to target the right visitors.
They work together: a proactive rule can start a Guided Tour, and a Chatbot Flow can be triggered by a proactive rule. Start with a single chat-message rule, then layer the others on.
Chatbot Flows script branching conversations inside your chat widget — greetings, keyword answers, guided paths — with no code. A flow is a set of connected steps (nodes): you ask a question, the visitor picks an option, and the flow sends them down the matching branch. Only Active flows run on your site.
When the path is predictable and you want it identical every time — booking steps, a returns wizard, a "which plan?" chooser, collecting an email before handoff. Flows never improvise.
When questions are open-ended — "does this work with my setup?". The Customer Chat AI answers from your trained knowledge. A flow can hand off to the AI (or a human) at any step.
Runs as the opening message when a visitor first opens the chat.
Runs when the visitor types a word you listed (e.g. refund, pricing, cancel).
Starts once contact details are shared, or a set time into the conversation.
Runs as the chat wraps up (e.g. a rating), or is started by a Proactive Message rule.
Click New Flow, name it, pick a trigger type (and keywords if Keyword), save. It starts switched off.
Click Edit Canvas — the visual builder where you add steps (messages, questions with buttons, branches) and connect them.
For each choice, draw a branch to the next step. End paths by resolving, handing to the AI, or transferring to a human.
Click Enable. The flow goes live in your widget immediately. Disable any time without losing the design.
No — they run side by side. Use flows for predictable paths and the AI for open-ended questions; a flow can always hand off to either.
Guided Tours are step-by-step walkthroughs that point at parts of your page with speech bubbles and a spotlight, guiding a visitor through onboarding or a new feature. They are fully themed and built in a visual designer with a live preview. A tour runs only when it is marked Active.
Each step targets an element and shows a bubble with your text. Add as many stops as the walkthrough needs and order them.
Start from a preset or design your own bubble — colours, gradients, spotlight — with a live preview on the right.
Optionally link a Proactive Rule so the tour auto-starts when that rule fires (e.g. a returning visitor's first dashboard visit).
A tour is eligible to run only while Active is ticked. Untick to take it down without deleting it.
Click New Guided Tour, name it, and optionally link a Proactive Rule so it starts automatically.
Add each stop: choose the element it points to and write the bubble text. Reorder until the path flows.
Pick a preset or style the bubble yourself. The live preview shows exactly what visitors see.
Tick Active and save. The tour can now run — on its own trigger, or started by a Proactive Message rule.
On their own trigger, or via a Proactive Message rule's "Guided tour" action — so you decide which visitors see which tour and when.
Conversion Goals turn an activity you care about — an inbound webhook event, a chat message, a ticket resolution — into a recorded conversion with full attribution. When the activity fires, OpsIQ logs a conversion against the contact and ties it back to the campaigns, segments and sources that led there.
Click New conversion goal. Choose the activity type that counts as a conversion, or use a Quick-start preset — Stripe purchase, Demo booked, Newsletter signup or Subscription started.
Give the conversion a name (purchase, demo_booked, signup) and optionally a fixed value and an attribution window.
Save and enable. From then on, whenever the activity fires, OpsIQ records a conversion against the contact with attribution to the preceding touches.
Any recorded activity — an inbound webhook event (payment.succeeded), a chat email capture, a ticket resolution — anything on the contact timeline.
Custom Dashboards let you build your own view of the metrics that matter, instead of relying only on the standard Overview. Add the widgets you want, drag and resize them into the layout you like, save it, and share it with your team.
Click Add Widget and pick the metric or chart you want to track. Repeat for everything you want on the board.
Drag widgets to reposition and drag their edges to resize. KPIs across the top, charts below — whatever reads best.
Save the dashboard and share it with your team so everyone watches the same numbers.
Outreach is OpsIQ's 1:1 and bulk email engine: compose a personalised message or run a campaign to a CRM segment, with reply detection, send-time control, open/click tracking, a tracking pixel and one-click unsubscribe. It is consent-gated and CAN-SPAM / CASL compliant.
Write a personalised 1:1 email with {tokens} from the contact and CRM.
Send to a CRM segment, scheduled or now, with caps and send-time control.
Reusable message templates for repeated outreach.
Reply detection records responses and auto-stops follow-ups to anyone who replied.
Choose a CRM segment (or upload a list). Suppression and consent are applied automatically.
Write the message with personalisation tokens; preview against a real contact.
Send now or at the best time. A tracking pixel records opens; clicks and replies are attributed.
Scheduled Reports email a chosen metric set to chosen recipients on a schedule — daily, weekly or monthly — as a printable, white-label PDF. Set it once and the report lands in inboxes automatically through the OpsIQ cron.
Pick the metric set (traffic, SEO, sales, support…) and a date range.
Daily, weekly or monthly, at a time you choose, to one or more email recipients.
Save and enable. The report generates and emails itself on schedule — no manual step.
The Privacy & Compliance page is the single place to show an auditor you handle personal data responsibly: the consent ledger, the admin access log, your data residency and retention policy, and a ready-to-print DPA — in one tabbed view. Only full administrators can open it.
A searchable record of every consent grant and revoke — by contact, channel (email, push, SMS, in-widget), action, source and date. Your proof a contact agreed (or withdrew).
Who on your team viewed, edited, exported, erased or messaged a contact's data — with the admin, contact, action, stated reason and IP. Exactly what a regulator asks for.
Declare where this workspace's data lives, set how many days activity rows are kept (0 = forever), and record your data controller and DPO.
A printable Data Processing Addendum that fills in your controller, address, DPO, residency and retention. Print or save as PDF.
Connectors link OpsIQ to external platforms: e-commerce stores, billing systems, payment processors, email providers, and custom APIs. Once connected, the AI can look up orders, customer data, subscriptions, and more — and take actions with your approval.
Web hosting billing: clients, services, tickets, invoices, domains, servers.
E-commerce: orders, customers, products, inventory, payments, webhooks, GDPR.
WordPress e-commerce: orders, customers, products, coupons, shipping.
Payments: charges, customers, invoices, subscriptions, payouts, disputes.
African payments: transactions, customers, plans, subscriptions.
E-commerce: orders, customers, products, categories, brands.
E-commerce: orders, customers, products, categories, inventory.
E-commerce: orders, customers, products, cart rules, carriers.
E-commerce: orders, customers, products, categories.
Legacy e-commerce: orders, customers, products.
Helpdesk: tickets, users, organizations, comments.
Local business: reviews, posts, Q&A, insights. Uses OAuth.
Email delivery: send transactional emails via Amazon SES.
Email delivery: send transactional emails via Postmark.
Email delivery: send transactional emails via Resend.
Email delivery: send emails and manage templates via SendGrid.
Email delivery: send emails via Mailgun.
Navigate to Connectors in the admin sidebar.
Browse the list or search. Connectors are grouped by category.
Click the connector and toggle Enable.
Fill in the required fields: API key, URL, tokens, etc. Each connector has specific requirements documented in its settings panel.
Click "Test Connection." A green result means the connector can reach the external platform. A red result shows the error.
Choose which features to enable: knowledge sync, webhook sync, AI actions. Not all connectors support all features.
The AI can look up customer data, orders, subscriptions, and more through the connector. When a customer chats, the AI has real platform data to reference.
Your connector's ContextProvider feeds the ticket-AI a read-only account summary of the ticket owner. Resolve the customer by <code>customer_id</code> when known; otherwise — on ticket auto-replies, where the ctx flag <code>customer_trust_email</code> is true — look the account up by <code>customer_email</code> (the verified ticket sender). Remote platforms expose a <code>/context/customer-by-email</code> endpoint and the connector calls it. This is what makes the AI quote a real balance/invoice instead of saying “please log in.” Honour the operator's data-class toggles (<code>customer_sections</code>, <code>customer_allow_credit</code>) and stay strictly read-only — never emit secrets.
Your connector's IdentityProvider feeds three universal ticket lookups: <code>lookupCustomerById()</code> (customer card), <code>lookupCustomerServices()</code> (product/service table), and <code>listCustomers($query,$limit)</code> — the "Search customer or type email" picker on the native Compose-New-Ticket form. Return rows of <code>{id, external_id, name, email, company, phone, platform, platform_label}</code>. <strong>Remote rule:</strong> if your customers live in another database or service, fetch them over your platform's API inside these methods — never read the other system's tables directly (in the OpsIQ DB they're empty placeholders, which is why a local read returns an empty picker). Gate the remote path on your configured API URL; keep a local read only as a same-DB fallback.
When inbound email becomes a native ticket, OpsIQ detects the sender by calling <code>lookupCustomerById(0, ['email'=>$sender])</code> across every connector; the first match "owns" the customer and its <code>platform</code> slug is stamped on the <code>ticket.created</code> event as <code>_pipe_owner_platform</code> (empty = guest). So a WHMCS-customer email opens a ticket in WHMCS (WHMCS numbering, mirrored as one), while an unknown sender stays a native guest. <strong>If your connector mirrors native tickets onto its own platform</strong>, your <code>ticket.created</code> handler must gate on this: <code>if (array_key_exists('_pipe_owner_platform',$payload) && $payload['_pipe_owner_platform'] !== '<your-slug>') return;</code> — the key is absent on non-piped (chat/manual/AI) tickets, so those mirror as before. Per-department control lives under Team & Departments → Open tickets via email (Anyone / Customers only / Off).
The AI can perform actions through the connector: create tickets, update orders, cancel subscriptions — all with your approval through the Trust Layer.
Connectors contribute platform-specific knowledge articles to the AI. Shopify connector adds e-commerce troubleshooting; WHMCS connector adds hosting knowledge.
If your platform hosts a knowledge base, declare a <code>kb_capabilities</code> block in <code>settings.json</code> (type <code>kb_source</code>) and your connector appears in the Help Center editor's <strong>Import KB → Import from a connected app</strong> dropdown whenever it is enabled and configured in the workspace. It is declarative — no method to implement: name a built-in <code>adapter</code> (<code>zendesk</code> for the exact-tree Help Center API, or <code>crawl</code> for any KB site) and a <code>base_template</code> such as <code>https://{subdomain}.zendesk.com</code>, which OpsIQ fills from your connector's stored settings. Credentials never leave the server; the browser only sends the connector slug. To support a new platform, add an <code>opsiq_kb_<platform>_manifest()</code> adapter and reference it by name. Full reference: <code>connectors/PLATFORM_CONNECTORS.md</code> §5.
Some connectors register webhooks with the external platform for real-time event notifications (new orders, ticket updates, etc.).
E-commerce connectors feed revenue data into the Sales dashboard for attribution and analytics.
Yes. You can have Shopify for orders and Stripe for payments in the same workspace. The AI merges data from all enabled connectors.
Use the Connector Builder to create a custom connector for any platform with a REST API. Or check the Marketplace for community-built connectors.
Yes. Connectors work identically in both cloud and self-hosted editions.
Every connector declares a set of actions: things the AI can do through the external platform. Actions are the bridge between "the AI says" and "the AI does."
"Cancel my subscription." The AI identifies this requires the "cancel_subscription" action.
Is this action allowed? Is it enabled? Does the current user have the right role?
The action needs a subscription ID. The AI looks it up from the customer context or asks the customer.
For destructive or high-impact actions, the AI asks for confirmation before executing.
The action calls the external platform API.
The AI reports the result to the customer.
Look up data without changing anything. Examples: list_orders, get_customer. Safe, no confirmation needed.
Create or modify data. Examples: create_ticket, update_order. May require confirmation.
Remove data. Examples: cancel_subscription. Always require confirmation.
Declared in actions.json — generic API calls defined declaratively.
Optional per-action "capability" field in actions.json — a connector-neutral class tag (e.g. "tickets" for a ticket read/lookup, "ticket_create" for opening a ticket). OpsIQ gates the whole class uniformly across every connector: ticket reads drop out of the AI when the operator turns off AI Ticket Awareness, and ticket creation is reserved for OpsIQ's own composer (so a connector's create tool can't duplicate it). Declare it and your connector conforms — OpsIQ never needs to know your action names. The Connector Builder exposes it as a Capability dropdown per action.
Implemented in PHP in the connector class. Full control over logic.
{
"lookup_order": {
"label": "Look up an order",
"description": "Find an order by ID or email",
"method": "GET",
"endpoint": "/admin/api/2024-01/orders.json",
"params": {
"order_id": { "type": "string", "label": "Order ID" },
"email": { "type": "email", "label": "Customer email" }
}
}
}The Connector Builder lets you create a complete connector for any platform with a REST API. No coding required for basic connectors.
The wizard has six steps, shown as tabs across the top. You can jump between them freely, and Save draft in the footer keeps your work at any point.
Name, slug, one-line description and an icon. The slug becomes the folder name under /connectors/, so it is lowercase with no spaces. This step is also where you tick the capabilities you want — each one adds working code to your connector, so you configure it rather than write it. The full list is below.
Pick the authentication pattern the platform uses and fill in its fields. Enter real test credentials here as well: they are used for the connection test on the next step, and they are not stored after you publish.
Enter the API base URL and a verify endpoint, then run a live connection test using the credentials from step 2. Do this before you define any actions — if the auth or the base URL is wrong, every action you write afterwards would fail for the same reason and you would be debugging the wrong thing.
Define each operation: name, HTTP method, endpoint path, parameters, response field mapping, and the reliability settings described below. You can run each action individually against the live API from this step.
Map the webhook events your platform sends onto OpsIQ events, so things that happen on your platform can start work in OpsIQ.
A summary of every action, trigger and capability you configured, plus any warnings. It also states plainly which advanced abilities the wizard cannot generate for you, so there are no surprises. Publishing writes the connector into /connectors/<slug>/ and it appears in the Connectors hub like a built-in one. Afterwards you can run a dry-run check or submit it to the marketplace.
Every action has the following settings. Only the first four are required; the rest control who may run it and how carefully.
The key is what the AI calls; the label is what a human reads. Keys are namespaced with your connector slug automatically, so two connectors can both have a "list_orders" without colliding.
GET, POST, PUT, PATCH or DELETE.
Appended to the base URL. Use {name} for placeholders, for example /opportunities/{id}/stage
For each parameter you also choose where it goes: <b>body</b>, <b>path</b>, <b>query</b> or <b>header</b>. A path parameter fills a {placeholder}; a query parameter is appended to the URL.
Say which field in the JSON response means what — which is the customer name, which is the email, which is the order total. This is what lets the AI read your platform's data rather than guess at it.
Your team (admin AI), signed-in customers (customer chat), both, or the writing assistant. Customer actions run only for the signed-in customer, and identity fields are locked to them automatically so one customer cannot ask about another.
A human must approve before the action runs.
Marks the action as deleting or changing state. Used to warn before it runs.
The admin must hold this permission for the AI to run the action on their behalf; full administrators always pass. Choose from manage_integrations, manage_general, manage_billing, manage_sales_bridge, manage_knowledge, manage_appearance, manage_behavior, manage_permissions, manage_dangerzone, reply_to_chats, escalate_to_ticket, send_agent_email — or none, meaning any authenticated admin.
Optional, but worth writing. It tells the AI when to reach for this action, for example "Use this when the admin asks to change a deal stage". A good hint is the difference between an action the AI uses correctly and one it ignores.
Optional class tag, either tickets (a ticket read or lookup) or ticket_create (opens a ticket). Tagging lets OpsIQ gate the whole class at once — for example, hiding ticket reads when an operator turns AI ticket awareness off.
Optional. For endpoints that return a list, the dotted path to the array, for example data or data.items
Optional. A body field whose presence means the call failed, for example error.message — useful for APIs that return HTTP 200 on failure.
Each action can be run against the live API from this step, so you can fix a wrong path or field name before anyone else uses it.
The Triggers step gives you a webhook URL to register with your platform, and verifies every request that arrives on it.
The header your platform signs with, for example X-Hub-Signature-256.
HMAC-SHA1, HMAC-SHA256 or HMAC-SHA512.
Simple — a plain hex or base64 digest. Or Stripe-style — the t=…,v1=… format used by Stripe and several platforms that copied it.
Optional. Some platforms prefix the digest, for example sha256=. Leave it empty if yours does not.
Map each event your platform sends onto the OpsIQ event it should raise.
If your platform has an API for managing its own webhooks, you do not have to register the URL by hand. Open Auto-register this webhook on enable and point it at two of the actions you defined in step 4:
Called when an admin enables the connector. You give the parameter name that carries the URL, optionally the one that carries the event list, and the path in the response that holds the new webhook's id.
Called when the connector is disabled, using the id captured above, so disabling the connector also stops your platform sending to a dead endpoint.
Tick a capability and the builder writes the code for it. You supply the endpoint and a field mapping; everything else is handled.
Operations the AI can run on your platform: list orders, find a customer, issue a refund.
Receive webhooks from your platform and turn them into OpsIQ events.
AI Triage and the reply Copilot on tickets that stay on your platform. Pick the endpoint that returns one ticket with its replies and map the fields; OpsIQ reads the thread through it.
Show a customer's mail on their timeline, send from a deal, and sync your address book into the OpsIQ CRM.
Page through your customer list and import each one as an OpsIQ contact. This is what the Import button runs.
Push live data from your platform into every AI conversation.
Work out which of your customers a visitor is, and show their profile and purchases on a ticket.
Feed revenue events and MRR figures into the OpsIQ dashboards.
Turn tickets and emails your platform delivers by webhook into OpsIQ support threads.
Post OpsIQ staff replies back to the platform the thread originated on.
Import your staff accounts into OpsIQ admin users.
Import your support departments into OpsIQ.
Let Promotion Studio target on facts you know about a person, and render your real products inside a promotion.
When OpsIQ blocks an IP, call your platform's own block API so the block is enforced there too.
Receive blocked-IP threat signals from OpsIQ.
Every connector you build ships a short reference describing what your platform is and what its actions do. The OpsIQ admin AI reads it whenever your connector is enabled, so it understands your platform rather than just seeing a list of action names. This is generated for you.
Your connector always gets a general-purpose call so the AI can reach an endpoint you did not define a typed action for. Use typed actions where you can; this is the escape hatch.
How many times to retry a failed call, with a backoff that doubles each attempt. Only timeouts, 429s and 5xx errors are retried; a 4xx is your request's fault and repeating it would just burn your rate limit.
A per-minute ceiling for this action, so a busy trigger cannot hammer your platform. Leave it at 0 for no limit.
How long to wait for your platform before giving up.
Require a human to approve before the action runs. Use it for anything destructive.
Step 2 offers seven patterns. Pick the one that matches your platform's API documentation — the generator emits the right HTTP client for it.
A static key sent as a fixed header, for example X-API-Key. You give the header name and the label the admin form should show. This sends the key as a header only; if your API expects the key as a query parameter instead, use Custom headers or a full-code connector.
Token sent as Authorization: Bearer <token>. Common for modern APIs.
Username and password, base64-encoded into the Authorization header. You choose the labels both fields show on the admin form.
A public key plus a secret, with the request body signed. You give the signature header, the key header and the timestamp header your platform expects.
Server-to-server, with no user interaction and no redirect. You give the token URL and an optional scope; the admin supplies a client ID and secret. Use this when your platform issues credentials to an application rather than to a person.
The authorization-code flow with PKCE, for Stripe, Shopify, Slack and QuickBooks-style apps. You give the authorization URL, the token URL and the scope. After publishing, the admin clicks Connect on the connector, authorizes on your platform, and OpsIQ stores the refresh token and mints access tokens on demand. The redirect URI to register in your OAuth app is shown on the connector's settings page.
A free-form JSON header map for anything the patterns above do not cover, for example {"X-API-Key":"{api_key}","X-Tenant":"{tenant_id}"}. Each placeholder becomes a field on the admin form.
Uses offset and limit parameters. Example: ?offset=100&limit=50.
Uses page number. Example: ?page=3&per_page=50.
Uses a cursor token from the previous response.
Uses the Link response header with rel="next".
<?php
namespace OpsIQ\Connectors\MyPlatform;
use OpsIQ\Connectors\AbstractConnector;
class MyPlatformConnector extends AbstractConnector
{
public function identifier(): string { return "myplatform"; }
public function label(): string { return "My Platform"; }
public function testConnection(): array
{
$resp = $this->http("GET", "/me");
return $resp["ok"]
? ["status" => "ok", "message" => "Connected"]
: ["status" => "error", "message" => $resp["error"]];
}
protected function runAction(string $action, array $params): array
{
return match ($action) {
"list_customers" => $this->http("GET", "/customers", [
"query" => ["page" => $params["page"] ?? 1],
]),
default => ["error" => "Unknown action: $action"],
};
}
}When your platform owns a ticket, mirror it into OpsIQ from your connector using the ticket ingest helper your connector inherits (create for a new ticket, add-reply for follow-ups). Emit and subscribe to the universal ticket events in your connector's subscribers() method — ticket.created, ticket.replied, ticket.updated, ticket.deleted. Auto-reply, escalation, SLA and CSAT all hook the same events, so you inherit them with no extra code.
A ticket is always for the customer, but the opening message may be written by staff — an agent opening a ticket on the customer's behalf, or reaching out proactively. Every helpdesk supports this, so it is part of the universal ticket.created contract rather than something each connector solves privately.
client (default) or admin. Omit it and the customer is assumed.
The staff id when author_type=admin. Never the customer id.
Agent or department display name. Falls back to "Staff".
The agent address. NEVER the customer one — a staff message under the customer e-mail is a reporting and audit error.
This is not cosmetic. Consumers read author_type to decide whose turn it is. A staff-written opening reported as client leaves auto-reply eligible to answer a message your own agent wrote, and makes your platform's echo of that opening look like a fresh customer message — which then lands as a duplicate and flips the ticket to answered. Map your platform's signal onto it in your connector's event normaliser, never in the consumers.
When an agent blocks a sender, that applies across every channel immediately — e-mail, chat, portal, widget and every connector twin — because the block is enforced centrally, before any connector sees the message. Your connector inherits it by doing nothing. There is no inbound work to implement and no way to get it wrong.
The optional half is outbound: telling your platform a customer was blocked here. Implement it only if your platform can express something equivalent.
Must include block_registry AND the class must implement BlockRegistryInterface. The flag is a claim; the interface is the guarantee — the registry requires both.
Receives kind (email|domain|ip), value, reason, admin_id, expires_at, site_key. Returns [result, detail?, remote_ref?].
Same payload, reversed. Must be idempotent — unblocking something never blocked is noop, not an error.
For reconciliation, so drift is reported rather than discovered by a customer. Returns [result, blocked?].
Applied on your platform. Include remote_ref when you have an id worth recording.
Already in the requested state.
Your platform has no equivalent concept. NOT an error.
You could have done it and it went wrong. Include detail; it lands in the audit trail.
Declining is a correct answer. Many platforms have no "block this sender" concept and offer only ACCOUNT suspension, which stops billing, logins and services rather than just support mail. Mapping "an agent blocked a spam address" onto "close this client's account" would let one mis-clicked block terminate a paying customer's services. If that is the only lever you have, return unsupported and leave it alone.
Declining also covers the case where the platform can do it but will not let you. Check for an API, not a feature: a lever that exists only in the platform's admin UI is a lever your connector does not have.
Read connectors/zendesk/connector.php alongside this. It is short, and every branch below is a decision you will face too.
PUT /api/v2/users/{id} with suspended=true. A suspended end user's new requests divert to Suspended Tickets instead of opening one — genuinely the same idea as blocking a sender, and it stops no billing because Zendesk end users have none. Returns the user id as remote_ref.
Zendesk CAN suspend a whole domain, but only through the suspend: keyword in the Admin Center blocklist textarea, and no API writes that field. The feature exists; the lever does not.
No ticket-level IP block exists.
Zendesk cannot pre-suspend an address it has never seen. Nothing to do is not a failure, and the local block still stops the ticket.
REFUSED. Suspending staff would lock a colleague out of the helpdesk. A staff address reaching a blocklist is far more likely to be a mistake than an instruction — check the role before you act on it.
That last row is the one worth copying. The contract cannot stop you mapping a block onto something destructive, so each connector has to decide for itself what it will refuse. Refusing loudly, with a detail an operator can read, beats obeying.
Report honestly. Notice how few of those rows are ok. That is the correct outcome, not a gap: a domain block still works — it is enforced in core the moment it is written — it simply cannot be mirrored outward. Returning ok for something you did not do would tell an operator the block reached your platform when it never left OpsIQ.
It is written BEFORE you are called and stands whatever you return. You cannot prevent an operator blocking someone.
A throw is treated as unsupported and recorded. Return failed with a detail instead so the reason survives.
The same block may be announced more than once.
Fan-out is synchronous with the operator click. Queue slow work rather than holding the request open.
Never re-block into OpsIQ from your handler. The registry skips the originating connector, but a second hop would loop.
Every announcement is written to opsiq_block_registry_log — operation, kind, value, reason, admin, origin, connector slug, result, detail and timestamp — so an operator can see a block arrived from your platform rather than from an agent, and what each connector did about it.
Hand an attachments array to create()/addReply(). Each item carries the file in one of three shapes:
Base64 of the RAW file bytes — use for files behind auth (most inbound cases). OpsIQ decodes it and re-hosts the file for you. The most portable shape.
A public https URL (e.g. a Zendesk content_url). Must be https; passes an SSRF / allowlist check before it is stored.
A stored file reference you already wrote (what the native composer uploader produces).
Items also accept filename, mime_type and size_bytes. Caps: 12 MB/file, ~24 MB total, 10 files per call. Most platform read APIs return attachment filenames only (not bytes), so the reliable inbound pattern is: read the bytes where your code already runs — inside the platform, or from the webhook payload — base64 them, and ship them as content_b64 on the event.
Outbound (OpsIQ → your platform). When an agent attaches a file, the ticket.replied payload gives you the reply's files as stored_path items under the OpsIQ file store. Read them and push in your platform's own format — and check the platform docs, the encoding matters: WHMCS wants base64_encode(json_encode([[name,data]])) (JSON, not serialize); Zendesk uploads raw bytes to /uploads.json for a token then sets comment.uploads; Gmail sends a multipart/mixed RFC-822 message base64url-encoded in raw; Amazon SES inbound attachments need the S3 receive action (the SNS path caps at 150 KB).
If your platform is a separate app or database (a hosted SaaS, a remote WHMCS, a different DB on the same server), never write its tables directly — a cross-database write lands on whatever connection is active and silently misses the real rows. Always: outbound = call the platform's signed REST API (replies, status, department, deletes); inbound = the platform POSTs a signed webhook, your handleWebhook() verifies it and fires the matching ticket.* event. To stop a mirror loop, tag what you push with an origin marker (e.g. mirror_origin:'opsiq'); the platform echoes it on the webhook and your inbound handler skips its own echoes. Check OpsIQEvents::isMirroredEvent() at the top of every handler.
A reply moves status automatically — an agent reply → answered, a customer reply → customer_reply; mirror that the same way. For explicit changes subscribe to ticket.updated and read payload['changes'] (e.g. {status:'closed'}).
Status: OpsIQ's canonical set is open, customer_reply, answered, in_progress, pending, closed. Keep a small two-way map to your platform's vocabulary, and make sure your column actually accepts the values you write (an ENUM missing a value gets silently dropped to empty by some databases). Department: the OpsIQ department id is meaningless on your side — never send it. Send the descriptor (name, email and slug) and resolve it against your own departments by e-mail → name → slug, so “Billing Department” lands in your Billing department instead of general. Trigger on both ticket.escalated and a department_id change in ticket.updated.
If the customer chose a specific product when opening the ticket, stamp a hint so the agent sees a focused service card (with a one-click "view other products" modal for the rest). Pass any of these to create():
Preferred — the service id exactly as it appears in your IdentityProvider::lookupCustomerServices() list (each service's id). Unambiguous match.
A platform service-code, e.g. WHMCS S396 (service 396) / D11 (domain 11). OpsIQ decomposes the optional type-letter + id and matches your service list, with a type guard so a domain code never matches a product of the same id. Expose a matching ref on each service for an exact match.
A domain name; matched as a substring against each service's detail / domain.
This is the source-level contract for building a connector that OpsIQ can discover, configure, validate, enable, execute, test, sign and distribute.
connectors/example_platform/
├── connector.php # connector implementation / entry point
├── settings.json # manifest + configuration schema
├── actions.json # optional action catalogue
├── README.md # setup, scopes, webhooks and limits
├── workflow_recipes.json # optional reusable recipes
├── knowledge.json # optional provider knowledge
├── profile.json # optional connector profile
├── src/ # optional client/provider classes
├── assets/ # optional admin assets
└── signature.json # signed release metadatasettings.json uses manifest version 1. The validator expects a slug, name, version, manifest_version, compatibility, description and capabilities. Compatibility declares opsiq_min and opsiq_max. Release status is one of production, beta, preview or template; capability maturity uses levels L1–L6. Legacy packages can be normalised with warnings, but a new package should pass without warnings.
slug(), name(), description(), capabilities(), settingsSchema(), testConnection().onEnable() and onDisable(). Disable must be non-destructive; data purge is a separate deliberate operation.handleWebhook(), poll(), backfill(). Poll handles incremental work; backfill imports history and must be safe to repeat.pushReply() where the platform accepts replies.registerActions() supplies executable action contracts.securityCapabilities(), syncBlockedIp(), removeBlockedIp(), testSecuritySync().The abstract base also offers optional chunked backfill, customer listing, email-account discovery/creation, subscribers, admin/customer detection, billing/context/identity providers, response extraction, retrying HTTP calls, JSON-path access, fuzzy ranking and pagination helpers.
Supported capability names include actions, triggers, context, identity, sales, billing, inbound_ticket, outbound_reply, sync_users, sync_departments, security_events, customer_import, security_native_block, promo_audience, promo_feed, email_mailbox, native_api, ai_brain_aware and ai_tickets. The builder provides API key, bearer, basic, HMAC, OAuth 2 client credentials, OAuth 2 authorisation code and custom authentication templates.
onEnable(), synchronises action definitions and can start the initial backfill.onDisable() and stops activity without deleting imported data.actions.json catalogue does not make an action usable until the connector is installed, enabled, configured and its connection is valid.Run PHP lint, manifest validation, connection tests, action schema tests, webhook signature/replay tests, workspace-isolation tests, backfill-repeat tests, disabled-connector tests and failure/retry tests. A marketplace package should then be signed and verified, include setup/permissions/webhook documentation, and state its true release status. The connector tile and manifest are the authoritative availability signal.
Step-by-step recipes for common connector setups.
Goal: Connect Shopify with automatic order sync and GDPR compliance.
1. Create a custom app in Shopify admin (Settings > Apps > Develop apps). 2. Grant scopes: read_orders, read_customers, read_products, write_customers. 3. Copy the Admin API access token. 4. In OpsIQ Connectors > Shopify, enter store domain and token. 5. Click Test Connection. 6. Enable Webhook sync — OpsIQ registers order/customer webhooks automatically. 7. GDPR webhooks register automatically (data_request, customers/redact, shop/redact). 8. Verify: place a test order, check it appears in OpsIQ Sales within 30 seconds.
Goal: Connect WHMCS so AI can look up clients, services, tickets, and invoices.
1. In WHMCS admin, Setup > Staff Management > Manage API Credentials. Create a key. 2. In OpsIQ Connectors > WHMCS, enter URL, API identifier, and secret. 3. Click Test Connection. 4. Install the WHMCS module addon for ticket relay. 5. Verify: ask the admin AI "Find client [email protected]" — WHMCS data should appear.
Goal: Track Stripe payments and let AI look up transactions.
1. In Stripe Dashboard > Developers > API keys, create a restricted key (read: Customers, Charges, Invoices, Subscriptions). 2. In OpsIQ Connectors > Stripe, enter the restricted key. 3. Test Connection. 4. Add webhook endpoint in Stripe: https://your-domain.com/connectors.php?slug=stripe 5. Select events: charge.succeeded, invoice.paid, customer.subscription.updated. 6. Copy the webhook signing secret to OpsIQ. 7. Verify: make a test payment, check it in OpsIQ Sales.
Goal: Connect GBP for review management, posts, and insights.
1. Create a Google Cloud project, enable Business Profile API. 2. Create OAuth 2.0 credentials (Web application). 3. Set redirect URI: https://your-domain.com/connectors.php?slug=google_business_profile 4. In OpsIQ, enter Client ID and Client Secret. 5. Click Authorize, sign in, grant access. 6. Select your business location. 7. Verify: reviews, posts, and insights should appear.
OpsIQ subsystems are open for extension. A connector can plug a new capability into the CRM or Site Intelligence with no changes to OpsIQ core, via the platform capability registry (\OpsIQ\Platform\CapabilityRegistry). Declare the capability, implement its small method set, ship the connector — OpsIQ discovers it and wires it into the UI, the Customer 360 timeline and the install prompts.
Add a <code><domain>_capabilities</code> array — best for no-code/marketplace connectors. Domains: <code>crm</code>, <code>site_intelligence</code>.
Implement <code>crmCapabilities()</code> / <code>siCapabilities()</code> returning the same descriptors.
Goal: make a calendar connector appear in the CRM Schedule-meeting panel.
"crm_capabilities": [ { "type": "calendar", "label": "Calendar (Acme)", "actions": ["createEvent","listEvents","freeBusy"], "requires_auth": true } ]
Goal: feed local-listing data into Site Intelligence.
"site_intelligence_capabilities": [ { "type": "local_listing", "actions": ["listLocations","listReviews","getInsights"], "requires_auth": true } ]
<code>createEvent($settings,$params)</code> → {success,event_id,html_link,meet_link}; <code>listEvents</code>; <code>freeBusy</code>. Driven by \OpsIQ\CRM\CalendarBridge.
<code>enrich($settings,$params)</code> → {success,traits}. Driven by \OpsIQ\CRM\EnrichmentBridge (also runs automatically on company create when a provider is connected).
<code>sendForSignature($settings,$params)</code> → {success,envelope_id,sign_url}. Driven by \OpsIQ\CRM\EsignBridge.
<code>listLocations</code>, <code>listReviews</code>, <code>getInsights</code>. Ships via the Google Business Profile connector.
<code>keywordRanks($settings,$params)</code> → {success,ranks}. Driven by \OpsIQ\SiteIntelligence\RankDataBridge. Ships: SerpApi connector.
<code>report($settings,$params)</code> (sync/export) + <code>ingest($settings,$events)</code> (import). Driven by \OpsIQ\Analytics\WebAnalyticsBridge. Ships: Google Analytics (GA4) connector — Data API for reports, Measurement Protocol for import. Analytics is open for connectors like CRM + Site Intelligence (domain <code>analytics</code> / <code>analytics_capabilities</code>).
<code>listPayments</code>, <code>customerPayments</code>, <code>createPaymentLink</code>. Driven by \OpsIQ\Payments\PaymentsBridge — show a customer's payments on the CRM timeline + send a payment link on a deal. Ships: Stripe Payments connector. Domain <code>payments</code> / <code>payments_capabilities</code>.
<code>listMessages</code> + <code>sendMessage</code>. Driven by \OpsIQ\Mailbox\MailboxBridge — a customer's emails on the Customer 360 timeline + send email from a deal. Ships: Gmail + Outlook/Microsoft 365 connectors. Domain <code>mailbox</code> / <code>mailbox_capabilities</code>.
All Google connectors share ONE app via \OpsIQ\OAuth\SharedGoogleApp — set <code>google_oauth_client_id</code>/<code>_secret</code> once and Calendar, Analytics, Business Profile + Gmail all work (the per-connector "(optional)" fields are overrides). They also share ONE redirect URI — <code><site>/connectors.php?oauth=callback</code> (no per-connector slug) — because Google exposes no API to add redirect URIs to an app; OpsIQ recovers the connector from the signed OAuth <code>state</code> nonce. Register that single URI once and every Google connector (present + future) authorizes with no further Console setup. Return <code>SharedGoogleApp::resolve(...)</code>’s <code>redirect_uri</code> from your <code>oauthParams()</code>; never hard-code a per-slug redirect. Microsoft connectors do the same via \OpsIQ\OAuth\SharedMicrosoftApp (<code>microsoft_oauth_*</code>).
The <strong>Import</strong> (historical backfill) and <strong>Import customers</strong> buttons show ONLY for connectors that actually implement them — detected by reflection, so there is no flag to declare. Override <code>backfill()</code> or <code>backfillChunk()</code> to earn the Import button; override <code>listCustomers()</code> (return {email, first_name, …} pages) to earn Import customers. Inherit the <code>AbstractConnector</code> no-op and the button stays hidden, so the UI never offers an import a connector cannot perform.
The CRM agent proposes and books meetings off deal signals (\OpsIQ\CRM\MeetingProposer → CalendarBridge), and auto-enriches new accounts — the “AI operates the CRM” loop, executed across connector buses.
fetchTicket())When a visitor references a ticket number in the client chat ("what's the status of T-000123?"), OpsIQ resolves it in two tiers. <strong>Tier 1</strong> reads the shared OpsIQ ticket store — every connector with <code>inbound_ticket</code> mirrors its tickets there, so synced tickets just work for ALL connectors with zero per-connector code. <strong>Tier 2</strong> is a live fallback for OLD tickets that were never mirrored: core calls an OPTIONAL <code>fetchTicket()</code> on each enabled connector that implements it. It is <strong>duck-typed</strong> (discovered by <code>method_exists()</code>, like <code>contextProviders()</code>/<code>audienceCatalog()</code>), so the signed <code>ConnectorInterface</code> is untouched and connectors that don't need it (email/chat sources that create tickets natively) simply omit it.
<?php
// connector.php — OPTIONAL universal client-chat ticket resolver (Tier 2).
// Only add it if your platform keeps its OWN ticket store that may hold tickets
// not yet mirrored into OpsIQ. Return the normalised array below, or null.
//
// $ref = ['mode' => 'id'|'tid', 'value' => string] (from extract_ticket_ref)
// $ctx = ['client_id' => int, 'client_email' => string] (VERIFIED visitor)
// $settings = your connector settings
public function fetchTicket(array $ref, array $ctx, array $settings): ?array
{
try {
$email = trim((string)($ctx['client_email'] ?? ''));
$cid = (int)($ctx['client_id'] ?? 0);
if ($cid <= 0 && $email === '') return null; // no verified identity → refuse
// 1) Resolve THIS visitor on your platform (by verified email/id).
// 2) List ONLY their tickets and match $ref against the ticket number/id.
// 3) Re-confirm ownership on the detail record before returning.
// On any miss/mismatch/error → return null (never leak).
return [
'tid' => 'T-000123', // the customer-facing reference
'id' => 123, // your internal id
'status' => 'open',
'subject' => '…',
'department' => 'Billing', // '' if your platform has none
'priority' => 'high',
'opened_at' => '2026-06-29 10:00:00',
'last_reply' => '2026-06-29 12:00:00',
'message' => 'opening message text',
'replies' => [ // oldest → newest; opening message may be first
['who' => 'You', 'date' => '…', 'message' => '…'],
['who' => 'Staff', 'date' => '…', 'message' => '…'],
],
];
} catch (\Throwable $e) {
error_log('[yourslug][fetchTicket] ' . $e->getMessage());
return null; // graceful degrade
}
}<code>CapabilityRegistry::providers($domain,$type)</code> lists providers with installed/enabled/ready status (filesystem-based; DB only for the enabled flag).
<code>CapabilityRegistry::active($domain,$type)</code> returns the first ready provider with its loaded connector and settings — bridges call this instead of hardcoding slugs.
Where a feature needs a capability that is not connected yet, the UI shows a one-click Connect/Install prompt — never a dead control.
See <code>connectors/PLATFORM_CONNECTORS.md</code> (CRM + Site Intelligence) and <code>connectors/CRM_CONNECTORS.md</code> in the package.
Signed-in customers can ask the customer chat AI about their OWN account — "my balance", "my orders", "did support answer my ticket" — and, if the business enables it, request refunds. All of it is powered by YOUR connector's declarations: no OpsIQ core changes, no platform hardcoding. The AI calls your lookups as native tools (it cannot invent one), identity is locked server-side to the verified customer, and results are automatically trimmed for the AI.
In your <code>actions.json</code>, a read-only action becomes customer-callable with ALL THREE: <code>"scope":"client"</code>, <code>"requires_confirmation":false</code>, <code>"is_destructive":false</code>. ⚠️ The value <code>customer</code> is NOT valid for scope — use <code>client</code> or <code>both</code>.
Name your identity parameter from the standard set — <code>client_id</code>, <code>customer_id</code>, <code>client_identifier</code>, <code>customer_identifier</code>, <code>account_id</code>, <code>email</code>, <code>customer_email</code>. OpsIQ locks and auto-fills it from the platform-verified signed-in customer; the AI never sees or chooses identity. A non-standard name fails closed (denied, never leaked).
Anything not scoped by that identity: fetch-by-bare-id (<code>order_id</code>/<code>ticket_id</code> alone), store-wide lists, free-text searches. Those are cross-customer leaks on any platform — keep them <code>"scope":"admin"</code>.
Two raw operations, policy-free: (1) a payment-verification read filterable by identity AND a transaction reference (your API must AND the filters), e.g. <code>whmcs_get_transactions</code>; (2) a refund execute action (<code>scope:"admin"</code>, <code>requires_confirmation:true</code>, <code>is_destructive:true</code> — never customer-exposed), e.g. <code>whmcs_refund_order</code>. The business picks both keys in Settings → Client Chat; OpsIQ re-verifies the payment server-side, applies the business's policy box (amount cap, age window, monthly limit), issues inside the box, and files a review ticket for everything else.
The Connector Builder's action editor has the same controls: set "Who can use it" to Signed-in customers, leave both write checkboxes off, and use a standard identity parameter name.
The "Customer Self-Service" entries in <code>connectors/whmcs/actions.json</code>, <code>woocommerce/actions.json</code>, <code>oscommerce/actions.json</code>; full contract in <code>connectors/README.md</code> §8.1–8.2 and <code>ConnectorInterface::registerActions()</code>.
Let the Promotion Studio target campaigns by LIVE client and billing state from YOUR platform — "has an unpaid invoice", "order pending", "subscription cancelled". Any connector (pre-installed, marketplace, or one you build yourself) joins by implementing two small methods. No OpsIQ core changes, no platform hardcoding: your connector appears as a platform in the studio Targeting tab with its own dropdown of facts, and several connectors can be enabled side by side.
Implement audienceCatalog() on your connector class. It returns a label and the list of keys your platform can answer. Each key shows up in the studio dropdown under your platform name.
Implement audienceFlags(array $identity). OpsIQ calls it at delivery time with the VERIFIED logged-in visitor identity (from the widget identity token — the same identity the client chat AI uses). Return a flat key => value map.
Enable the connector. The Promotion Studio Targeting tab grows a "Client & billing" section listing your platform; admins build rules like "unpaid_invoices >= 1"; every fact also becomes a personalization variable such as {unpaid_invoices}.
public function audienceCatalog(): array
{
return ['label' => 'My Billing Platform', 'keys' => [
['key' => 'unpaid_invoices', 'label' => 'Unpaid invoices (count)', 'type' => 'number'],
['key' => 'pending_orders', 'label' => 'Pending orders (count)', 'type' => 'number'],
['key' => 'client_status', 'label' => 'Client status', 'type' => 'text'],
]];
}
public function audienceFlags(array $identity): array
{
// $identity = verified token payload: ['email' => ..., 'id' => ..., 'name' => ...]
$email = strtolower(trim((string)($identity['email'] ?? '')));
if ($email === '') return [];
$client = $this->api('GetClient', ['email' => $email]); // your platform call
if (!$client) return []; // unknown client -> return [] (fails CLOSED)
return [
'unpaid_invoices' => (int)$client['unpaid_count'],
'pending_orders' => (int)$client['pending_orders'],
'client_status' => (string)$client['status'],
];
}Add <code>'promo_audience'</code> to your connector's <code>capabilities()</code> array so the platform and marketplace list it as audience-capable (discovery also works from method presence, but declaring it is the clean signal). The pre-installed <code>connectors/whmcs</code> connector is the full reference implementation.
Lowercase snake_case, 2-60 chars, matching ^[a-z0-9_]+$. Types: number, text, bool. Undeclared keys returned by audienceFlags() are still usable in rules, but declare everything you support so admins can see it.
The $identity array comes from the signed widget identity token issued by the customer site at login — never from client-supplied fields. If your platform cannot resolve the identity, return an empty array.
On ANY error (API down, client not found, timeout) return []. A rule whose fact is missing makes the campaign NOT show. A billing-targeted popup must never reach the wrong person; OpsIQ enforces this in the rule matcher too.
audienceFlags() is called at most once per connector per page view, and ONLY when a published campaign actually uses your platform's rules. Anonymous visitors cost you zero calls. Keep it to 1-2 API calls; OpsIQ memoizes within the request.
Admins combine your keys with =, !=, >, >=, <, <=, contains, is set, is empty — you only supply values.
Every fact doubles as a variable in campaign copy: {unpaid_invoices}, {client_status}, plus the built-ins {first_name} and {client_email}. Values are sanitized before rendering.
Declarative (JSON-only) marketplace connectors get promo_audience for FREE: if your connector declares customers/users, invoices, orders or subscriptions resources, OpsIQ derives customer_status, unpaid_invoices, amount_due, pending_orders, active_subscriptions and subscription_status from your existing search actions. No code at all.
Add an audience_capabilities object — {"label": "My Platform", "keys": [{"key": "open_carts", "label": "Open carts", "type": "number"}]} — to control exactly which facts (and labels) appear in the studio dropdown.
Connectors > your connector > Enable. Then open any campaign's Targeting tab — your platform must appear under "Client & billing (connectors)".
Add: Your platform · a key · >= · 1 and set "Logged-in clients only". Save and publish.
Log in to your platform as a client who matches the rule (popup must show) and as one who does not (popup must stay hidden). Logged-out visitors never match logged-in rules.
Beyond the built-in triggers (delay, scroll, exit-intent, idle, click, pages-viewed, chat / ticket activity), you can fire a campaign from anywhere in your site or connector front-end — after a successful checkout step, when a cart is abandoned, when your SPA changes route, etc.
Set the campaign trigger to <strong>Custom</strong> with an event name (e.g. <code>checkout_failed</code>), then call the global hook from your page:
// Any campaign whose Custom trigger event matches will be evaluated
// (it still respects targeting, frequency caps and audience rules).
window.OpsIQPromo && window.OpsIQPromo.fire('checkout_failed');The Marketplace lets you discover, install, and publish connectors built by the community.
Go to Connectors > Marketplace. Search by name, category, or platform.
Check the description, supported actions, and ratings.
Click Install. The connector package downloads and installs to marketplace_connectors/.
Enable it in Connectors, enter API credentials, and test.
Use the Connector Builder. Verify all actions work.
Generate profile.json, actions.json, knowledge.json, and PHP files.
Run php tools/resign_all_connectors.php to sign the package.
Upload through the marketplace submission flow for review.
Ship with OpsIQ. Maintained by the OpsIQ team.
Installed from marketplace. Can be declarative or custom PHP.
The API and Webhook Builder lets you create custom API endpoints and webhook listeners inside OpsIQ without writing raw PHP.
Run when something happens in OpsIQ, or when an external system posts a matching event to your webhook URL. Type the event name, for example ticket.created.
Use one of the built-in interval events as the event name: cron.minutely, cron.hourly or cron.daily. The Actions and Triggers screen has buttons for these, so you do not have to remember them.
Run when an admin asks the AI for it in plain language. Add the phrase that should match, for example find customer.
An external system sends an HTTP POST to your OpsIQ webhook URL with an event name. Any trigger listening for that event fires.
An event trigger can carry conditions, so it only runs when the incoming data matches. Without conditions it runs every time the event fires.
A path into the event data. Use a dot for nested values, for example customer.email.
Choose whether every rule must match, or just one of them.
equals, does not equal, contains, does not contain, starts with, is one of, is not one of. Matching ignores capitals.
greater than, at least, less than, at most. Numbers written as text still compare correctly.
is present, is missing, is empty, is not empty.
Event ticket.created, with one rule: priority equals urgent.
The action runs for urgent tickets and is skipped for everything else. No filtering needed at your end.
Send HTTP POST to an external URL with a custom payload.
Create a support ticket with specified department, subject, and message.
Send an email with a custom template.
Update a customer, ticket, or CRM record.
Execute a connector action.
Ready-to-use trigger-action recipes for common automations.
Monitoring system sends POST when server goes down.
Trigger: Inbound webhook Condition: payload.status = "down" Action: Create ticket — Dept: Technical, Subject: "Server {{payload.hostname}} down", Priority: Urgent
Slack message every time a customer places an order over $100.
Trigger: Event (order.completed) Condition: order.amount > 100 Action: Send webhook to Slack incoming webhook URL with order details
Summary email to management every Monday at 9am.
Trigger: Schedule (Monday 09:00) Action: Send email with last week's stats pulled from the REST API
If a ticket has no reply for 24 hours, assign to team lead.
Trigger: Schedule (every hour) Condition: ticket.status = "Open" AND ticket.last_reply_age > 86400 Action: Assign to team-lead, add internal note
This walkthrough creates a simple connector action from scratch.
Before you start, decide which one you need. Both are configured in the same screen.
The action sends an HTTP request to your platform. Choose this when OpsIQ needs to reach out and do something on your side.
The action runs something an installed connector already provides. Choose this when the work is already available and you just want to trigger it. You pick the connector action from a list, so there is nothing to wire up.
Example: "lookup_product" — look up a product by name, return price, stock, and description.
Add the action definition:
{
"lookup_product": {
"label": "Look up a product",
"description": "Find a product by name",
"method": "GET",
"endpoint": "/api/products/search",
"params": {
"name": { "type": "string", "label": "Product name", "required": true }
},
"response_mapping": {
"id": "$.data.id",
"name": "$.data.name",
"price": "$.data.price",
"stock": "$.data.stock_quantity"
}
}
}In connector settings, use dry-run to test with sample data.
Ask admin AI: "Look up product Widget Pro." The AI should find and execute the action.
Inbound endpoints accept signed JSON from another system and turn it into deduplicated contact activity that can feed timelines, scoring and workflows.
Create one source per sending system. OpsIQ generates a URL/token and secret. Send JSON with Content-Type: application/json and sign the exact raw body using HMAC-SHA256:
X-OpsIQ-Inbound-Signature: sha256=<hex hmac>Map JSON paths for the contact email, event name and optional event ID. The event ID is the preferred deduplication key. Accepted activity enters the contact timeline and can affect lead scoring, segments, lifecycle rules and workflows.
Reject invalid signatures before parsing business data. Rotate a secret by updating both sides inside a controlled window; do not log the secret.
Outbound webhook administration creates signed subscriptions, tests endpoints and monitors asynchronous delivery and retry health.
Full administrators create an HTTPS endpoint, choose specific events or the wildcard, store the secret safely and send a test. OpsIQ signs the raw payload with HMAC-SHA256:
X-OpsIQ-Signature: sha256=<hex hmac>Verify the signature before deserialising or acting. Delivery is asynchronous and automatically retried after approximately 1 minute, 5 minutes, 30 minutes, 2 hours and 12 hours. Use the recent-delivery rail to inspect status, response code and attempt count.
Your receiver must be idempotent because any asynchronous system can redeliver. Return a 2xx response only after durable acceptance.
Webhooks let external systems notify OpsIQ when something happens (inbound), and let OpsIQ notify external systems when something happens internally (outbound).
Go to Settings > Webhooks. Add an endpoint URL and select events to receive.
ticket.created, ticket.replied, chat.started, chat.message, order.completed, crm.deal.stage_changed, crm.contact.created, etc.
Failed deliveries retry with exponential backoff. After 15 consecutive failures, the endpoint is auto-disabled.
Every request includes an HMAC signature in X-OpsIQ-Signature. Always verify.
OpsIQ will not send to private/internal IPs (127.x, 10.x, 172.16.x, 192.168.x, 169.254.169.254).
Webhooks are delivered via the job queue, not blocking OpsIQ operations.
Each event has a unique event_id that stays the same across retries for deduplication.
<?php
$payload = file_get_contents("php://input");
$signature = $_SERVER["HTTP_X_OPSIQ_SIGNATURE"] ?? "";
$secret = "opsiq_whsec_your_secret";
$expected = hash_hmac("sha256", $payload, $secret);
if (!hash_equals($expected, $signature)) {
http_response_code(401);
exit(json_encode(["error" => "Invalid signature"]));
}
$event = json_decode($payload, true);
// Process the event...
http_response_code(200);
echo json_encode(["received" => true]);const crypto = require("crypto");
app.post("/opsiq-webhook", (req, res) => {
const payload = JSON.stringify(req.body);
const sig = req.headers["x-opsiq-signature"];
const expected = crypto.createHmac("sha256", "opsiq_whsec_xxx")
.update(payload).digest("hex");
if (!crypto.timingSafeEqual(Buffer.from(expected), Buffer.from(sig || "")))
return res.status(401).json({ error: "Invalid signature" });
console.log("Webhook:", req.body);
res.json({ received: true });
});Use ngrok to expose your local server. Set the webhook endpoint to your ngrok URL.
OpsIQ retries with exponential backoff. If your endpoint recovers within the retry window, no events are lost.
API keys authenticate requests to the OpsIQ REST API. Each key has scoped permissions.
Navigate to API key management.
Give it a descriptive name.
Give the key only what it needs. Scopes come in read and write pairs per area, so a reporting integration can be granted reads and nothing else.
Pick the narrowest set that does the job. admin satisfies every scope below, so use it only when you genuinely need broad access.
Contacts, companies, deals, activities, segments, outreach, conversions, lifecycle and CRM webhooks.
List and read tickets and their messages; reply, update, open and run bulk queue operations.
Read live-chat conversations and message history; send messages, assign and change status.
Knowledge base entries.
Saved reply templates.
See which connectors are installed, what they can do, and test their credentials.
See your actions, triggers and their run history.
Departments available for routing.
Notifications for a named operator.
Report sections, CSV exports and scheduled report definitions.
Monitored sites and their SEO reports.
Signed download links for attachments.
Status page components and incidents.
AI writing helpers: draft, summarise, translate, rewrite, compose, analyse.
Everything, including a small number of sensitive actions that are deliberately not reachable by any narrow scope: erasing a contact, changing privacy settings, and reading the access log.
The key may call anything its scopes allow.
Every write is refused, whatever the scopes say. The safest choice for reporting, dashboards and anything that should never change data.
You list the exact actions the key may call. Anything else is refused, even within an allowed scope.
Shown once. Starts with "opq_". Store it securely.
The OpsIQ widget is a JavaScript snippet you add to your website for tracking, AI chat, ticket creation, and customer identification.
<script src="https://your-opsiq-domain.com/widget.php" data-site-key="site_abc123" async></script>
Paste before the closing </body> tag. The script loads asynchronously and does not block page rendering.
Required. Your site key from Connected Sites.
Optional. "chat" (default), "tickets", or "both".
Optional. "bottom-right" (default) or "bottom-left".
Optional. Primary color hex code.
Optional. Initial greeting message.
Optional. UI language: "en", "es", "fr", "de", "pt", "ar", etc.
Optional. Server-generated signed identity token.
// Manually notify OpsIQ of a route change (for custom routers)
window.OpsIQ = window.OpsIQ || [];
window.OpsIQ.push(["pageview", {
url: window.location.href,
title: document.title,
}]);script-src 'self' https://your-opsiq-domain.com; connect-src 'self' https://your-opsiq-domain.com; frame-src https://your-opsiq-domain.com;
No. Under 15 KB gzipped, loads async after page content.
Use data-color for primary color. Contact support for CSS override options.
Platform-specific installation guides.
Add to all pages of a WordPress site.
Option 1: Edit theme footer.php, paste snippet before </body>. Option 2: Use "Insert Headers and Footers" plugin > Scripts in Footer. Option 3: Use a child theme footer.php to survive theme updates.
Add to your Shopify store.
Online Store > Themes > Edit Code > Layout > theme.liquid. Paste before </body>. Save.
Add to a React or Next.js app.
In Next.js, use next/script in app/layout.tsx: import Script from "next/script"; <Script src="https://your-opsiq.com/widget.php" data-site-key="site_abc" strategy="lazyOnload" />
Add to your WHMCS client area.
Setup > General Settings > Other > Global Footer Content. Paste the snippet. Or edit your template footer.tpl.
Add via GTM.
Create Custom HTML tag, paste snippet. Trigger: All Pages. Publish.
Identity connects anonymous visitors to known customers. When identified, OpsIQ links browsing history, chats, and tickets to the customer profile.
window.OpsIQ = window.OpsIQ || [];
window.OpsIQ.push(["identify", {
email: "[email protected]",
name: "Jane Smith",
customer_id: "cust_1001",
plan: "pro",
company: "Acme Inc",
}]);<?php
$secret = "opsiq_whsec_your_secret";
$payload = json_encode([
"site_key" => "site_abc123",
"customer" => [
"id" => $user->id,
"email" => $user->email,
"name" => $user->name,
],
"iat" => time(),
"exp" => time() + 900,
]);
$token = base64_encode($payload) . "." . hash_hmac("sha256", $payload, $secret);They become anonymous until they log in again. OpsIQ links the new anonymous record when re-identified.
OpsIQ provides SDK packages and plugins for common platforms.
Composer package: identity token generation, action API helpers, webhook signature verification.
Browser-side: tracking, identity, custom events, chat control.
Widget injection, identity from WP user sessions, WooCommerce sync.
Full integration: widget, ticket relay, identity, admin AI.
App embed with automatic customer identity.
composer require opsiq/sdk
use OpsIQ\SDK\Identity;
$token = Identity::createToken(
siteKey: "site_abc123",
secret: "opsiq_whsec_your_secret",
customer: ["id" => "cust_1001", "email" => "[email protected]", "name" => "Jane"]
);
// Call any catalog action over the action API
$ch = curl_init("https://your-opsiq.com/api/v1.php");
curl_setopt_array($ch, [
CURLOPT_POST => true,
CURLOPT_RETURNTRANSFER => true,
CURLOPT_HTTPHEADER => [
"Authorization: Bearer opq_your_key",
"Content-Type: application/json",
],
CURLOPT_POSTFIELDS => json_encode([
"action" => "crm.contacts.list",
"status" => "lead",
"limit" => 25,
]),
]);
$contacts = json_decode(curl_exec($ch), true);The OpsIQ REST API lets you programmatically access all platform features.
Send the key as Authorization: Bearer opq_... or X-OpsIQ-Key: opq_.... Use one authentication header, not both.
POST https://your-opsiq-domain.com/api/v1.php
Authorization: Bearer opq_your_api_key
# Alternative header: X-OpsIQ-Key: opq_your_api_key
Content-Type: application/json
{
"action": "endpoint.name",
"site_key": "site_abc123"
}The API is action-based: send an action name in the body. POST {"action":"meta.actions"} (no key required) for the authoritative, per-install list of every action and the scope it needs.
List CRM contacts (scope crm.read). Params: q, status, source, tag, lead score, limit.
Resolve-or-create a contact (scope crm.write). Params: email plus fields, tags, notes.
List CRM deals (scope crm.read). Params: status, stage_id, company_id, owner_admin_id, pipeline_id, limit.
Record a lifecycle/business activity for a contact (scope crm.write). Params: email or contact_id, event, value, currency.
Record a conversion (scope crm.write). Params: contact_id or email, conversion_type, value, currency.
Poll the CRM event stream (scope crm.read). Params: limit.
List tickets (scope tickets.read). Params: page, per_page, sort, order, status, priority, department_id, assigned_admin_id, client_email, search.
One ticket and its messages (scope tickets.read). Params: ticket_id or ticket_number, include_messages, messages_limit.
Post a reply to a ticket (scope tickets.write). Params: ticket_id or ticket_number, body, author_type, author_name, internal.
Update a ticket (scope tickets.write). Params: ticket_id or ticket_number, status, priority, subject, department_id, assigned_admin_id.
Open a support ticket (scope tickets.write). Params: customer_email, customer_name, subject, body, priority.
Run a bulk queue operation (scope tickets.write). Params: operation, ids.
List live-chat conversations (scope conversations.read). Params: page, per_page, status, assigned_admin_id, client_email, search.
One conversation and its messages (scope conversations.read). Params: conversation_id or thread_key, include_messages.
Paged message history (scope conversations.read). Params: conversation_id, page, per_page, include_internal.
Send a message into a conversation (scope conversations.write). Params: conversation_id, message, sender_type, internal.
Assign, re-status or re-title a conversation (scope conversations.write). Params: conversation_id, status, assigned_admin_id, subject.
See which connectors are installed, what each one can do, and whether its credentials work (scope connectors.read).
The actions your Actions builder will run (scope automations.read).
What will fire and on which conditions (scope automations.read).
Execution history for debugging an integration (scope automations.read). Params: status, action_key.
Departments you can route a ticket to (scope departments.read).
Notifications for one operator (scope notifications.read). Params: admin_id (required), unread_only.
A report section as CSV (scope analytics.read). Params: section, preset, from, to. Call analytics.sections.list for valid values.
Monitored sites and their SEO reports (scope site_intelligence.read).
A short-lived signed download link for an attachment (scope files.read). Params: reference, ttl.
Every list action uses the same parameters, so you learn them once.
Page number starting at 1, and page size. Default 50, maximum 200. limit and offset are also accepted if you prefer them.
Field to sort by, and asc or desc (default desc). Each action accepts a fixed set of sort fields; anything else is ignored rather than erroring.
Send filters as an object, or pass a filter as a top-level parameter. Unknown filters are ignored.
items, page, per_page, count, total and has_more. Keep requesting pages until has_more is false.
POST /api/v1.php
Authorization: Bearer opq_your_api_key
{
"action": "tickets.list",
"status": "open",
"priority": "urgent",
"sort": "created_at",
"order": "desc",
"page": 1,
"per_page": 50
}Every failure returns the same shape, with a stable code you can branch on. Do not match on the human-readable error text, which may be reworded.
{
"success": false,
"code": "missing_scope",
"error": "API key missing required scope. Needed one of: tickets.read, admin",
"request_id": "req_8f2c1a9d4b7e"
}No key was sent, or the key is wrong or expired. HTTP 401.
The key is valid but lacks the scope this action needs. HTTP 403.
The key is read-only, or this action is not in its allow-list. HTTP 403.
No such action. Call meta.actions for the current list. HTTP 400.
A required parameter is absent, or a value is not accepted. HTTP 400.
The record does not exist, or is not in the workspace this key can see. HTTP 404.
The hourly limit for this key is used up. HTTP 429.
Something failed on our side. The request_id identifies the attempt if you need to report it. HTTP 500.
If a write times out you cannot tell whether it landed. Send an idempotency_key and a repeat of the same call returns the original result instead of doing the work twice.
{
"action": "tickets.reply",
"ticket_id": 4192,
"body": "Thanks, that is now fixed.",
"idempotency_key": "your-unique-id-for-this-attempt"
}A key is remembered for 24 hours against the same action and API key.
The original response, plus idempotent_replay: true so you can tell it was a replay.
Only successful writes are remembered. If a call failed, sending it again with the same key runs it again, which is what you want.
Reads ignore the key. Repeating a read is harmless.
Add dry_run: true to any write. Your key, scopes and parameters are all checked exactly as normal, then the call stops before anything is created or changed.
{
"action": "tickets.update",
"ticket_id": 4192,
"status": "closed",
"dry_run": true
}
{
"success": true,
"dry_run": true,
"would_execute": true,
"message": "Authorised and validated. No write was performed because dry_run was set."
}A configurable hourly limit per API key (default 1000 requests per hour). Set it per key under Settings, API Keys.
HTTP 429 with code rate_limited, plus a Retry-After header telling you how many seconds to wait. The limit resets at the top of the next clock hour.
Every response carries X-RateLimit-Limit, X-RateLimit-Remaining and X-RateLimit-Reset (a Unix timestamp). Watch Remaining and slow down before you run out rather than waiting to be refused.
A 429 means wait and try again. A 401 means the key itself is wrong or expired, so retrying will not help until you fix the key.
Repeated failed-auth attempts from one IP are throttled separately. Valid keys never trip this.
Back off when Remaining gets low, and honour Retry-After after a 429 instead of retrying immediately.
v1. Every response carries X-OpsIQ-API-Version and X-OpsIQ-API-Status so you can confirm what you are talking to.
If a version is ever retired, responses gain standard Deprecation and Sunset headers, and meta.actions reports deprecated and sunset_at. Check those in your monitoring and you will never be surprised.
Existing action names, their parameters and their error codes. New actions and new optional fields may be added at any time, so ignore fields you do not recognise rather than failing on them.
This is the whole loop: find work, read it, act on it, and confirm what you did.
# 1. Discover what this install offers and what each action needs.
POST /api/v1.php
{"action":"meta.actions"}
# 2. Find the urgent open tickets.
POST /api/v1.php
Authorization: Bearer opq_your_api_key
{"action":"tickets.list","status":"open","priority":"urgent","per_page":25}
# 3. Read one, including its conversation history.
{"action":"tickets.get","ticket_id":4192}
# 4. Reply, safely enough to retry if the network drops.
{"action":"tickets.reply","ticket_id":4192,
"body":"We have issued the refund.","idempotency_key":"reply-4192-attempt-1"}
# 5. Close it.
{"action":"tickets.update","ticket_id":4192,"status":"closed"}
# 6. Page through anything larger: keep going until has_more is false.
{"action":"tickets.list","page":2,"per_page":50}This list is generated from the running catalogue, so it always matches the version you are on. meta.actions returns the same information as JSON.
Read. Scope: analytics.read, admin.
Read. Scope: analytics.read, admin.
Read. Scope: analytics.read, admin.
Read. Scope: automations.read, admin.
Read. Scope: automations.read, admin.
Read. Scope: automations.read, admin.
Write. Scope: chat.write, admin.
Read. Scope: chat.read, admin.
Write. Scope: chat.write, admin.
Read. Scope: connectors.read, admin.
Read. Scope: connectors.read, admin.
Read. Scope: connectors.read, admin.
Read. Scope: conversations.read, admin.
Read. Scope: conversations.read, admin.
Read. Scope: conversations.read, admin.
Write. Scope: conversations.write, admin.
Write. Scope: conversations.write, admin.
Read. Scope: admin.
Read. Scope: crm.read.
Write. Scope: crm.write.
Read. Scope: crm.read.
Read. Scope: crm.read.
Write. Scope: admin.
Read. Scope: crm.read.
Read. Scope: crm.read.
Write. Scope: crm.write.
Read. Scope: crm.read.
Read. Scope: crm.read.
Write. Scope: crm.write.
Write. Scope: crm.write.
Write. Scope: crm.write.
Write. Scope: crm.write.
Write. Scope: crm.write.
Write. Scope: crm.write.
Write. Scope: crm.write.
Read. Scope: crm.read.
Write. Scope: crm.write.
Read. Scope: crm.read.
Write. Scope: crm.write.
Write. Scope: crm.write.
Write. Scope: crm.write.
Read. Scope: crm.read.
Read. Scope: crm.read.
Write. Scope: crm.write.
Read. Scope: crm.read.
Read. Scope: crm.read.
Read. Scope: crm.read.
Write. Scope: crm.write.
Read. Scope: crm.read.
Write. Scope: crm.write.
Write. Scope: crm.write.
Write. Scope: crm.write.
Write. Scope: crm.write.
Read. Scope: crm.read.
Write. Scope: crm.write.
Write. Scope: crm.write.
Read. Scope: crm.read.
Write. Scope: crm.write.
Write. Scope: crm.write.
Write. Scope: crm.write.
Write. Scope: crm.write.
Read. Scope: crm.read.
Read. Scope: crm.read.
Write. Scope: crm.write.
Read. Scope: crm.read.
Write. Scope: admin.
Write. Scope: crm.write.
Read. Scope: crm.read.
Read. Scope: crm.read.
Read. Scope: crm.read.
Write. Scope: crm.write.
Write. Scope: crm.write.
Write. Scope: crm.write.
Read. Scope: crm.read.
Write. Scope: crm.write.
Read. Scope: crm.read.
Write. Scope: crm.write.
Write. Scope: crm.write.
Write. Scope: crm.write.
Read. Scope: crm.read.
Write. Scope: crm.write.
Read. Scope: departments.read, admin.
Read. Scope: files.read, admin.
Write. Scope: kb.write, admin.
Write. Scope: kb.write, admin.
Write. Scope: kb.write, admin.
Read. Scope: kb.read, read, admin.
Write. Scope: kb.write, admin.
Write. Scope: kb.write, admin.
Read. Scope: no scope required.
Read. Scope: no scope required.
Read. Scope: notifications.read, admin.
Read. Scope: site_intelligence.read, admin.
Read. Scope: site_intelligence.read, admin.
Write. Scope: status.write, admin.
Write. Scope: status.write, admin.
Write. Scope: status.write, admin.
Write. Scope: tickets.write, admin.
Write. Scope: tickets.write, admin.
Write. Scope: tickets.write, admin.
Read. Scope: tickets.read, admin.
Read. Scope: tickets.read, admin.
Write. Scope: tickets.write, admin.
Write. Scope: tickets.write, admin.
Write. Scope: tickets.write, admin.
Write. Scope: tickets.write, admin.
Write. Scope: tickets.write, admin.
Write. Scope: tickets.write, admin.
Read. Scope: tickets.read, admin.
Write. Scope: tickets.write, admin.
Write. Scope: tickets.write, admin.
Write. Scope: tickets.write, admin.
Write. Scope: tickets.write, admin.
Write. Scope: todos.write, admin.
Write. Scope: writing.
Write. Scope: writing.
Write. Scope: writing.
Write. Scope: writing.
Write. Scope: writing.
Write. Scope: writing.
Yes. The machine-readable OpenAPI 3.0 document is at /api/v1/openapi.php (import it into Postman, Insomnia, Stoplight, or a code generator). Human-readable developer docs are at /api/docs/.
Send site_key in the request body to choose a workspace. A key issued for one workspace stays pinned to it; a key with admin context can select any workspace it is permitted to see, and is refused for any it is not.
Any action marked as a read. Set the permissions mode on the key to read_only and every write is refused with read_only_key, so a reporting integration cannot change anything even by mistake.
The key is scoped to a workspace. Records in other workspaces return not_found rather than a permission error, so the API never confirms whether something exists elsewhere.
Send custom events from your server or website to OpsIQ for analytics, experiments, CRM triggers, and webhook conditions.
window.OpsIQ = window.OpsIQ || [];
window.OpsIQ.push(["track", "button_clicked", {
button: "signup",
page: "/pricing",
variant: "green"
}]);Server-side, record activities and conversions against a contact through the action API. Use crm.activities.record for lifecycle/business activity and crm.conversions.record for conversions. There is no events.send action; the inbound CRM event receiver lives at /v1/inbound.php?t=<token> for third-party systems that push events in.
POST /api/v1.php
Authorization: Bearer opq_your_key
Content-Type: application/json
{
"action": "crm.activities.record",
"site_key": "site_abc123",
"email": "[email protected]",
"event": "subscription.upgraded",
"value": 50,
"currency": "USD"
}"order.completed" not "orderCompleted".
"button.clicked" not "button.click".
"pricing.plan_selected" not "user_action".
Understanding the data flow helps you debug integrations and optimize performance.
The JavaScript widget sends page views, clicks, and chat messages to /beacon.php.
Events stored in opsiq_events with site_key, event type, customer, and properties.
Session builder groups page views into sessions (via cron).
When identified, anonymous records are linked to the customer profile.
Connectors pull data from external platforms and store in the connector cache.
AI assembles context from: visitor history, customer profile, connector data, KB, and prompt.
Internal event fires (ticket.created, chat.message, order.completed).
Event placed in the async job queue.
Dispatcher signs and sends to all subscribed endpoints.
Failed deliveries retried with exponential backoff.
Settings control how OpsIQ behaves across tracking, AI, tickets, security, and branding.
Your company name. Used in AI responses, emails, and widget header.
Your business industry. Helps the AI understand context.
All analytics and timestamps use this timezone.
Currency for revenue reporting.
Business hours for auto-reply scheduling and SLA timers.
Inactivity time before a session ends. Default: 30 minutes.
How long a visitor is "live" after last page load. Default: 90 seconds.
Replace last IP octet with 0 for GDPR compliance.
Respect DNT browser headers.
IP addresses to skip in tracking. Add your team IPs.
URL patterns to skip. Example: /admin/*
How long to keep visitor data. Default: 90 days.
Appears in admin header, emails, and widget.
Accent color throughout the admin UI and widget.
Browser tab icon for your OpsIQ admin.
Invite team members and assign roles to control who can access what.
Full access including license, billing, and danger zone. One per installation.
Full access except license and danger zone. Can manage admins.
Limited to assigned departments. Tickets, chats, and analytics only.
Navigate to Team in the sidebar.
Enter the team member's email.
Choose Owner, Full Admin, or Agent.
For Agents, select accessible departments.
They receive an email with a setup link.
How your support team is performing. Only accessible to Owners and Full Admins.
Average time from ticket creation to resolution per agent.
Average time to first staff reply. Most important SLA metric.
Average number of replies needed to resolve a ticket.
Open ticket count per agent right now.
Breakdown of chat vs. tickets vs. email per agent.
Customer satisfaction scores per agent vs. team average.
Current period vs. previous period comparison.
When each agent handles the most tickets. Use for shift planning.
OpsIQ tracks admin online status via a 60-second heartbeat. Admins not seen in 5 minutes are offline. The status widget appears in the admin navigation.
Sites (workspaces) let you manage multiple websites from one OpsIQ installation. Each has its own tracking, AI, connectors, and team access.
Navigate to Connected Sites.
Enter name, domain, and description.
Each site gets a unique key (site_xxx) for the widget snippet.
Each site has its own AI, connectors, and settings.
Depends on your license plan. Check your license details.
Your OpsIQ license controls features, site limits, and AI credit balance.
Enter in Settings > License. Validated against the OpsIQ licence server.
OpsIQ normally reads its signed local licence cache and refreshes validation when required. Use Force remote check when you need an immediate result; Diagnostics identifies DNS, HTTPS or server-time problems.
Determines feature availability and limits.
Maximum number of workspaces.
If using Managed AI, your credit balance is shown here.
Renew before expiration to avoid service interruption.
Tracking, tickets, and CRM continue. AI features pause until renewal. Data is preserved.
Each administrator can maintain their own identity, working preferences, password and two-factor authentication.
Profile settings belong to the signed-in administrator and do not change workspace branding. Use a unique password and enable two-factor authentication. Changing an email or authentication setting may require re-verification or a fresh sign-in.
Notification preferences control which operational events reach an administrator in the product or by email.
Preferences cover ticket assignment, customer replies, SLA warnings, chat handoff and AI approval events. Keep urgent operational signals enabled somewhere; turning off email does not necessarily turn off the in-app notification.
The Security Overview is the first screen for assessing blocks, login attacks, threat sessions, security trends and event integrity.
Use the headline cards for triage, then open the related Audit, Failed Logins, Blocked IPs or JavaScript Errors page for record-level work. The 14-day trend is more useful than a single spike. High-threat sessions should be investigated against visitor, request and authentication evidence before a permanent block is applied.
The audit log records who did what, when, to which object, at what risk level and whether it succeeded.
Filter by action, risk and actor; export CSV or JSON for an investigation. The hash chain helps detect missing or altered audit rows, but it does not replace protected backups and restricted database access. Treat audit exports as sensitive because they may identify users, records and security actions.
Diagnostics turns subsystem checks into plain-English health results and direct next actions.
The page checks Site Speed, rank/keyword/backlink/local data, AI and action gateway, tickets, crawler/audits, integration/event intake, background tasks, CRM automations, licence, widget and related product services. It intentionally reports operator-readable causes rather than exposing raw secrets.
After a change, Refresh and confirm the affected subsystem returns healthy. A green diagnostics page is necessary but not sufficient for launch; complete the Pre-launch checklist as well.
OpsIQ includes multiple security layers to protect your admin, data, and API access.
Email/password with optional 2FA.
Temporary lock after too many failed attempts.
Block IPs or CIDR ranges from accessing tracked sites.
Behavior-based threat scores. High-threat visitors can be auto-blocked.
Required for all connections. HTTP redirects to HTTPS.
Strict-Transport-Security prevents downgrade attacks.
All admin forms include CSRF tokens.
All user input is escaped. CSP headers restrict script execution.
Outbound webhooks block private/internal IP addresses.
For all admin accounts.
Minimum 12 characters, mixed.
Remove access for former team members.
Check weekly for suspicious patterns.
Apply updates promptly for security patches.
Role-Based Access Control lets you control what each team member can do.
Everything including license, danger zone, and installation management.
Everything except license and danger zone. Can manage team.
Respond to assigned-department tickets and chats, view analytics. No settings, team, or security access.
Not currently. Three built-in roles plus department assignment gives granular control.
Manage IP addresses blocked from accessing your tracked sites.
Example: 203.0.113.42 or 203.0.113.0/24
Document why: spam, abuse, scraping, etc.
If customers share a corporate IP or VPN, blocking that IP blocks everyone on it. Use narrow blocks.
All unsuccessful login attempts to your OpsIQ admin.
When the attempt happened.
Email address used.
Source of the attempt.
Geographic location.
Wrong password, locked, or not found.
Password guessing. Ensure 2FA is enabled.
Automated attack. Block the source IP.
If your team is local, foreign attempts are likely attacks.
An opt-in analyst layer over the Security pages: verdicts, incident briefs, a daily digest, a weekly posture check, Ask Security, and 0-token block policies. It never runs in the visitor request path.
Every Security page records what happened, but reading it still takes judgment. The AI security analyst adds that judgment in place, and it is built to spend almost nothing: it never sits in the beacon or login path, deterministic rules settle the obvious cases with no AI call, verdicts are cached, and the analyst's best output is a deterministic rule that then enforces forever at zero tokens. Turn features on in Settings → Security; all default off.
Blocked IPs and Failed Logins rows show a benign / suspicious / hostile chip with a confidence score. Explain opens the reasoning, the facts used, and one-click Block / Release actions. Cached 24h, so re-viewing an IP is free.
Related events cluster into one incident (login wave, URL scan, auto-block burst) with a plain-English timeline and one recommended next step. Ongoing attacks fold into the same incident.
One recap per day: yesterday versus your baseline, blocks, incidents, and the single thing to do. Quiet days send a deterministic all-clear at zero tokens.
A Monday hardening checklist (2FA gaps, lockout threshold versus real attack volume, alerts off) with a severity, effort and exact fix per item.
A question box on the Security Overview that answers from your live security state. Deep server questions are handed to the admin engineer chat rather than guessed.
The analyst proposes concrete block rules as Approve / Dismiss cards; approved rules enforce at zero tokens with a live hit counter. Optional autopilot approves only the safest rules automatically.
Master switch — enables verdict chips, Explain, and Ask Security. Only ambiguous IPs reach the model, batched into one call.
Clusters events into incidents and writes each new story. Runs on the existing hourly security cron.
One security recap per day on the bell and the Overview. Quiet days cost nothing.
Monday hardening checklist on the Overview. The checklist is fully deterministic.
Lets the analyst activate its safest proposals unattended: reversible time-limited blocks only, single IPs or ranges no wider than /24, never shared/mobile/private, capped 10 per hour, always audited and undoable.
OpsIQ uses one product scheduler to run email polling, ticket automation, analytics rollups, retention, AI queues, CRM work, reports and connector jobs.
Add this single line to the crontab for the same server and PHP environment that runs OpsIQ:
* * * * * php /path/to/opsiq/opsiq/cron/all.php >> /tmp/opsiq-cron.log 2>&1
Replace /path/to/opsiq with the installation root. Use the full PHP CLI path when the host requires one, for example /usr/bin/php8.4.
Do not use the top-level /cron.php for product work. In this repository that runner belongs to a separate non-product service outside this handbook. The OpsIQ product runner is opsiq/cron/all.php.
Processes due auto-replies, escalations, approvals and other queued product work.
Polls configured mailboxes where polling is the selected intake method.
Assembles sessions and refreshes scheduled aggregates when they are due.
Runs scoring, lifecycle, workflow, campaign and scheduled-report tasks that are enabled.
Performs due cleanup, archive, expiry and integrity work.
Dispatches connector cron subscribers and polling only for installed, enabled and configured connectors.
/tmp/opsiq-cron.log or the log path you selected when a manual run fails.Keep connector handlers short, idempotent and workspace-scoped. Batch long synchronisations and store a cursor/checkpoint so the next run can continue safely.
A hosted OpsIQ deployment uses the same product screens and workflows while the hosting operator manages the server layer. Commercial sign-up, plans, billing and account-portal procedures are outside this handbook.
Do not assume that a managed service means every product integration is healthy. Confirm the latest backup evidence and retention policy with the hosting operator, and use OpsIQ Diagnostics after changes to DNS, email, AI providers, websites or connectors.
This section covers product operations only. Commercial onboarding, subscriptions, billing and customer-account administration remain outside the OpsIQ handbook.
A self-hosted OpsIQ installation runs on infrastructure you control. You are responsible for the runtime, database, TLS, product cron, backups, updates and monitoring.
PHP 8.4 or newer. Required extensions: curl, pdo_mysql, mbstring, json, xml, openssl, zip, intl and gd.
MySQL 5.7+ or MariaDB 10.3+ using InnoDB. Create a dedicated database and least-privileged application user.
Apache or Nginx configured for the OpsIQ front controller, static assets and required rewrites.
Required for admin sessions, widget traffic, API keys, webhooks and provider calls.
Allow DNS and HTTPS access to the AI, email, connector and licence endpoints you actually enable.
The web and CLI users must be able to read application files and write only to the documented runtime/storage locations.
Run the OpsIQ product scheduler every minute with the same application configuration and database access as the web process.
* * * * * php /path/to/opsiq/opsiq/cron/all.php >> /tmp/opsiq-cron.log 2>&1
Checklist and best practices for running OpsIQ in production.
SSL installed, HTTP redirects, HSTS enabled.
Verify via diagnostics page.
Chat as a customer. Verify accuracy and tone.
On all website pages. Verify with Live Feed.
Automated daily database and file backups.
PHP error log enabled and monitored.
All members have accounts with correct roles.
At least 5-10 articles.
Run the source gates and complete real product checks before a release or first production launch. A command being available is not evidence that it passed in your environment.
Run the repository checks from the installation root and keep their complete output with the release record.
php tools/preflight.php
php tools/release_gate.php
The preflight checks the PHP 8.4 runtime and required extensions, performs a PHP syntax sweep, runs the AI evaluation ship gate and full PHPUnit suite, and runs PHPStan/Node checks when those tools are installed. The extended gate also includes smoke, clean-install/schema and browser end-to-end checks when their environment is configured.
A skipped check is not a pass. Record why it was unavailable, install the missing dependency or run it in the correct CI/staging environment before release.
Sign in as each intended role; verify allowed and denied pages/actions.
Load a real site in a private browser; verify beacon traffic, Live Feed, visitor and session records.
Ask representative questions; verify knowledge grounding, safety, escalation and AI history.
Create, assign, reply, escalate, close and reopen through every enabled intake path.
Test outbound delivery and each configured inbound mailbox or webhook path.
Create a controlled contact/deal/conversion and verify dashboards, funnels, scoring and reports.
Run connection tests, one read, one safe write/dry run and webhook/polling paths for each enabled connector.
Verify narrow scopes, rate limits, idempotent retries, signatures and duplicate handling.
Run the unified product scheduler and confirm all required heartbeats in Cron Setup/Diagnostics.
Verify 2FA, CSRF/session controls, audit entries, key revocation and workspace isolation.
Restore a current backup into a separate test environment and verify data and configuration.
Confirm logs/alerts, preserve request IDs and prove the rollback procedure.
Browser push notifications — including OpsIQ Push Campaigns — need a service worker served from the SAME ORIGIN as the page. The OpsIQ widget loads from your OpsIQ deployment, so a third-party site cannot register OpsIQ's worker directly. Host one small file on the site's own domain to enable push there. Campaigns deliver through BOTH supported methods.
Browsers only let a page register a service worker from its own origin. The widget loads cross-origin from your OpsIQ deployment, so it cannot install the push worker on a customer site — until you host the worker on that site's origin, the widget silently skips push (no error, no subscription). You have two options below; OpsIQ Push Campaigns publish through both.
1. Download the worker from your OpsIQ deployment: https://YOUR-OPSIQ/opsiq/opsiq-push-sw.js
2. Upload it to your web root so it is reachable at https://yoursite.com/opsiq-push-sw.js. The widget auto-detects this path on the page's own origin and registers it — no snippet change needed. (It uses a narrow scope, so it never replaces a service worker you already run.)
Optional: if you must host it at a non-standard path, point the widget at it before the snippet — it is only accepted when same-origin:
<script>window.OpsIQ=window.OpsIQ||{};window.OpsIQ.pushSwUrl="/custom/opsiq-push-sw.js";</script>3. In OpsIQ, generate VAPID keys (Settings → Client Chat AI → Advanced → Web Push). The widget then asks visitors for notification permission and subscribes them.
/opsiq-push-sw.js returns HTTP 200 (purge the cache if it was 404 before you uploaded — a cached 404 will block registration).1. In OpsIQ settings set the push provider to Pusher Beams and enter your Beams Instance ID and Secret Key.
2. Host the Beams service worker on the site's root as /service-worker.js. Copy https://YOUR-OPSIQ/opsiq/pusher-beams-service-worker.js, or merge this line into your existing root worker:
importScripts("https://js.pusher.com/beams/service-worker.js");3. That's it. The widget auto-loads the Beams SDK and subscribes each visitor to the per-site interest opsiq-site-<your-site-key>. No other wiring needed.
Push Campaigns publish through both methods at once: visitors on the self-hosted VAPID worker get the encrypted Web Push, and visitors subscribed via Pusher Beams get the Beams publish. Each visitor is reached through whichever they subscribed to, so you can run same-origin VAPID on some sites and cross-origin Beams on others.
HTTPS is required for service workers and Web Push (localhost is exempt for testing). The admin chat inbox already uses the same Beams interest mechanism for operator alerts.
Common issues and how to fix them.
Check PHP error logs. Causes: old PHP (need 8.4+), missing extension, DB connection failed, wrong file permissions.
Check: widget installed? Cron running? Date range correct? IP excluded?
Check: provider configured? API key valid (use Test connection)? Budget exhausted? Workspace AI config correct?
Check: knowledge base up to date? Prompt clear? Connector lookups returning correct data? Review conversation in AI History.
Depends on: provider API speed, context size (large KB = more tokens = slower), network latency. Try a faster model.
Check: script in page source? JS errors in console? CSP blocking? Ad blocker?
GeoIP not configured. Use Cloudflare (automatic) or MaxMind GeoLite2 database.
Check: mailbox connected? Cron running? App password changed? IMAP settings correct?
Check: enabled? Department allowed? AI configured? Cron running? Human replied first?
Check: credentials correct? Key expired? API accessible from server? Base URL correct?
Check: API key scopes? Webhook sync enabled? Cron running?
Check: database performance (SHOW PROCESSLIST), PHP memory (increase to 1GB), enable opcache, reduce retention period.
Use this focused check when the widget loads incorrectly or Live Feed stays empty.
When the widget is not logging visits, verify the site key, load the page in a private browser, check the browser Network/Console panels for blocked beacon or widget requests, review Content-Security-Policy and ad blockers, confirm the site is active in OpsIQ, and open Diagnostics. A visible chat bubble does not by itself prove that the analytics beacon was accepted.
Definitions for terms used throughout OpsIQ.
Something the AI can do through a connector or the platform.
A team member with limited access, or an AI agent.
Bring Your Own Key — you provide your own AI API key.
A plugin integrating an external platform with OpsIQ.
Customer Satisfaction score.
A revenue opportunity in the CRM.
A ticket category for routing and access control.
Moving a ticket to a different department or from AI to human.
Visitor location detection from IP address.
Transferring a conversation from AI to human.
Signed token identifying a website visitor.
Articles the AI uses to answer questions.
A visitor showing buying intent.
0-100 number indicating conversion likelihood.
Where a customer is: Lead, Prospect, Customer, At Risk, Churned.
AI provided as part of your OpsIQ plan.
Visual board of deal stages.
Role-Based Access Control.
A series of page views within a timeout window.
Unique identifier for a tracked website.
Service Level Agreement — target response times.
Unit of AI text. Roughly 4 characters = 1 token.
Safety system controlling AI autonomy.
HTTP callback for event notifications.
JavaScript code embedded on your website.
A site with its own tracking, AI, connectors, and team. Same as "site."
If this documentation does not answer your question:
Click "Ask OpsIQ" in the admin. Has access to this documentation and your platform data.
Include: what you tried, what happened, what you expected, and any error messages.
Email support with your license key and issue description.
Which admin page (URL or name).
Exact setting name and current value.
Exact error message with codes.
What you did, step by step.
What should happen vs what did happen.