Token arrives
Identity rides in on a signed token with the chat. No "verify your email" loop, no copy-pasting order numbers.
Client Chat AI · live · grounded · agentic
Chat that knows the customer.
A polished AI live-chat widget that recognises the visitor, reasons through the request, looks up their real account, answers from your knowledge base, and acts safely when asked. It escalates to your team with full context, and only opens a ticket when something genuinely needs follow-up. AI keeps talking while your visitor waits for a human. Light and dark, fully white-label.
Live account lookupsReasons before it repliesGrounded + safe actionsHandoff carries full context
Recognises logged-in visitors
Resolves before a human is needed
Live
Premium live chat
One widget, every conversation.
40customer languages available
7flow trigger types
13deterministic flow nodes
LiveAI continuity during handoff
🆔 Identity-aware context
Every reply starts with who they are.
The moment a known customer opens the chat, the inbox shows a live visitor card with their name, login status, last order, return frequency and whole journey. Identity rides in on a signed token, so the AI shares the same context and never asks "what's your order number?" again.
Live recognitionIdentified
SourceSigned identity token
Card builds
Name, account status, last order, visit count and the live journey assemble on the visitor card in real time.
AI reads it
The agent and the AI share the same context, so nothing resets and the customer is never asked twice.
Grounded, then handed off
Answers from your docs.
The AI answers from your knowledge base, FAQs and AI Instructions, not the open internet. When it can't answer confidently, the question doesn't vanish: it lands in Pending KB suggestions as a drafted article you Approve & Publish in one click. And when confidence drops or a visitor asks for a person, the chat escalates into your agent inbox with the full transcript, visitor card and journey already attached. It becomes a native ticket only when it needs follow-up.
Grounded, not guessing. It refuses to invent facts outside your content.
KB gap capture. Every miss becomes a drafted suggestion with the real question that triggered it.
Department routing to a ticket. Escalate to the right team with the transcript, attachments and SLA timers carried over.
71%resolved without a human
1-clickApprove & Publish
Groundedanswers from your KB only
Auto-draftevery miss → KB article
Full contexttranscript + journey on handoff
Agentic by design
The chat that thinks
The chat that thinks
before it replies.
Not a canned line. Every turn, it reasons through the request, pulls the customer’s real account with read-only tools, grounds the answer in your knowledge, and acts only when it’s safe. It never invents a fact, a balance or a ticket number.
🔒read-only tools
💰balance $360.00
📚grounded in your KB
🎫#TKT-000418 opened
01
02
03
04
05
06
Recognise
Verifies who they are from the signed identity token, so there’s no “verify your email” loop.
Plan the turn
Reads the question and decides exactly what it needs to check to answer it completely.
Look up the account
Pulls their real data with read-only tools: balance, invoices, services, domains and their tickets.
Ground the answer
Combines their live data with your knowledge base, FAQs and AI Instructions. Your facts, never the open web.
Act, safely
When asked, opens a routed ticket in the customer’s voice, files a refund for human review, or hands off.
Never fabricate
Read-only by default, so it never claims an action it didn’t take. Every capability is yours to toggle.
Read-only by default
Grounded, never invented
Every capability you toggle
Real data · safe actions
Their account, not just your docs.
“What’s my balance?” “When does my domain renew?” “Show my open tickets.” A signed-in customer gets a real answer, pulled live from their own account with read-only tools. It won’t guess, and it never touches anyone else’s data. When they want something done, the AI acts safely: it opens a routed ticket in their own voice, files a refund for a human to approve, or routes them to the right team. Then it quotes a ticket number and status the customer can actually look up.
-
Live, identity-scoped lookupsInvoices, services, domains, subscriptions, balance and their own tickets. Read-only, and only ever their own.
-
Composed on their behalfA ticket the AI opens reads like the customer wrote it, and routes to the right department by your rules.
-
Human-in-the-loop for moneyRefunds and sensitive changes are filed for a human to approve, never run by the AI itself.
🔒Read-only · only their data
✓1 ticket · one number
“What’s my balance and open tickets?”
signed-in · verifiedLive account lookupread-only
Wallet balance$360.00
Business planrenews 12 Aug
Open tickets0
Ticket opened · on their behalf
“Hi Billing team, I’d like a discount on my Business plan…”
Live datatheir real account, read-only
First-persontickets written as the customer
Approval-gatedmoney moves need a human
Opt-in image understanding
Send the screenshot. OpsIQ sees the problem.
A customer can paste the images they are looking at instead of describing every pixel. Client Chat inspects only the images on that current message, connects the visual evidence to approved knowledge and permitted account context, then answers once, with the original files kept private.
AM
Amina M.Signed in · Growth plan
LIVEMy payment went through, but this screen is still blocking me. What is wrong?
app.yourcompany.com
WORKSPACE STATUSActivation pending
Your payment was received. Provisioning has not completed.
RECEIPTPayment successfulInvoice #1042 · Paid
▧ checkout-error.webp▧ receipt.png2 images
IQCurrent messagevisual context
Image evidenceActivation pending
seenAccount checkInvoice 1042 paid
verifiedApproved guidanceProvisioning recovery
groundedOPSIQ AIGrounded visual reply
answeringInspecting 2 current-message images
Your screenshots show that payment succeeded, but workspace activation is still pending. I also confirmed invoice #1042 is paid.
You do not need to pay again. Try the activation refresh once; if provisioning remains pending, I can create a routed support ticket with these screenshots attached.
Image inspectedAccount checkedGuidance grounded
Paste or attachError screenshots, product photos and interface states.
Up to three at onceOne multi-image message becomes one coherent visual answer.
Current turn onlyPast images are never resent on later messages.
Private and fail-softPrivate storage; text-only reply if vision is unavailable.
SupportedJPEG, PNG, GIF and WebP · up to 4 MB each
ControlledOff by default · requires a vision-capable model
ScopedOnly the visitor’s newest message enters the vision call
Human handoff without the dead air
It doesn’t go quiet when you hand off.
The AI writes the handoff in the visitor’s language, stays useful while the queue moves, and gives control to the teammate who joins. If nobody arrives, it safely returns the conversation to AI, with context intact.
Account accessSarah Kim · signed in
LIVE1AI2Handoff3Jess
Sarah · Korean
결제는 완료됐지만 계정이 아직 활성화되지 않았습니다.
OpsIQ AI · 한국어
결제는 확인됐고, 활성화가 아직 처리 중입니다. 지원팀을 연결하는 동안 제가 계속 도와드릴게요.
Bringing in Billing1 min waiting · AI remains available
Sarah
지금 제가 할 수 있는 게 있나요?
OpsIQ AI · wait companion
You don’t need to retry payment. I’ve checked the paid invoice and kept the activation details ready for the teammate.
↺
No teammate joined in timeConversation safely returned to AI · write actions remain off
AI is still here…
The handoff line is written, not stored.
OpsIQ composes the transition from the live conversation in the visitor’s language instead of sending a generic system phrase.
The AI remains a useful companion.
It knows the minutes waited and acknowledgment count, stops repeating the wait after two mentions, and cannot run write actions while a person is being brought in.
If nobody joins, the AI comes back.
A widget timer, boot poll, and cron backstop can return a timed-out handoff to AI with a contextual comeback, not a stranded queue.
Follow-up that respects silence.
Optional follow-ups space out exponentially, stop at your configured limit, skip acknowledgments and active flows, and fail silently instead of pestering the visitor.
⚡ Engage on purpose
Reach out at the right moment.
Behavioural automation that fires a chat message, banner, email, guided tour or webhook based on what the visitor is doing. Plus deterministic, node-based flows for the moments you need certainty.
Proactive & welcome rules
Behavioural triggers, cooled down.
Fire on page URL, time on page, returning visitor or scroll depth. Cooldowns and per-visitor caps keep it tasteful; built-in A/B testing finds the message that works.
- Behavioural triggers: Page URL, Time on Page, Returning visitor, Scroll depth.
- Seven action types: chat, banner, email, fire webhook, assign dept, guided tour or push.
- Cooldowns, caps & A/B: max fires per visitor, cooldown hours and a live A/B variant reply rate.
Live
🧩 Chatbot flow builder
Deterministic flows where you need certainty.
For lead qualification, triage or guided onboarding, build node-based flows with branching buttons. Start them on a greeting, keyword, known details, delay, chat end, proactive rule or manually, and test the route in a live Flow Preview, no code.
- Seven trigger types: Greeting, Keyword, After Details, Delay, On Chat End, Proactive Rule and Manual.
- 13 runtime node types: messages, questions, branches, conditions, collection, knowledge, AI, email, tickets, delays and escalation.
- Variables & once-per-conversation: interpolate {{visitor_name}}; re-fire controls never nag returning visitors.
Live
One AI · every channel · every language
Meet every customer in the same governed inbox.
Website chat, WhatsApp, SMS, Messenger, Instagram, Telegram, LINE and X become normal Client Chat conversations in the Admin Chat Inbox. The source, external identity, attachments and reply rules travel with every thread.
□OpsIQ Inbox
Nabtech⌄
Admin⌄
LH
Layla Hassan● Client · WhatsApp · Arabic
✓ Join chat⊘ Decline↺ Return to AI× Close▰ Archive
Type in English, customer receives Arabic automatically
General3 selected
Jess: I’m joining Layla’s billing conversation now.
Settings → Client Chat
Every customer moment has a control behind it.
Shape the identity, language, behaviour, support journey, knowledge and delivery layer without rebuilding the widget. The preview follows the setting you touch so teams can configure with confidence.
IdentityRepliesLanguagesAccount data
Live previewFollow customer languageAutomatic detection on
LIVEAIQueueHumanAI
Business hours · department handoff · timeout return · offline form · Help Center entry · rating & survey
RealtimePollingPusherAblyFirebase
PushBeamsOneSignalVAPID
Channels7 connector routesAI per channel
Names, greetings, suggestions, launcher art, avatars, 55 theme recipes, glass, complete color tokens, geometry and placement.
Reply pace, typewriter, status cues, natural follow-ups, long-chat memory, account lookups, links, prices, tickets and refund-request guardrails.
Auto-assignment, working hours, offline forms, ticket awareness, Help Center in chat, ratings, surveys, feedback, retention, crawling, embeddings and verified FAQs.
Uploads and limits, realtime plus fallback polling, background push, PWA alerts, SLA, saved views, search, ticket inspector and per-channel AI policies.
Speak now · richer voice experiences next
Your customer can talk when typing is too slow.
Browser dictation is live in the widget today: tap the mic, speak, review the transcribed message, and send. Full browser calling, premium speech-to-text and text-to-speech, and an avatar experience are clearly marked roadmap, not presented as shipped features.
IQ
Support AI Online now
Hi Maya. How can I help?
My renewal date is wrong on the invoice.
Listening00:08
LIVE TRANSCRIPT
“The invoice shows the wrong renewal date, and I need the corrected copy…”
Review before sending
Browser voice input
SpeechRecognition turns speech into editable text inside the existing composer. If the browser cannot support it, the control stays out of the way.
Premium STT + TTS
Provider-grade transcription and spoken replies with workspace voice controls.
Browser voice and video
Real-time calls that retain customer identity, conversation context, and the human handoff.
AI avatar experience
An optional visual concierge built on the same governed Client Chat brain.
The trust layer underneath the bubble
Fast where it should be. Strict where it has to be.
The widget is only the surface. Behind it, OpsIQ validates where requests came from, normalizes provider streaming, preserves feedback through deletion choices, and keeps every brand detail configurable.
CLIENT CHATIQgoverned runtime
Copied embeds do not inherit your access.
request→registered domain→allowed
Registered site domains, the admin-managed allowlist, and the server’s own origin form the request boundary. Unknown origins receive no access header.
One live reply contract across providers.
AnthropicOpenAIGrokGemini
Visible deltas stream through one runtime, usage is accounted for, partial replies never double-send, and safe synchronous fallback remains available.
Deleting a chat cannot quietly erase a rating.
Keep recordKeep personErase
Choose what remains for accountability. Scheduled retention and Forget Visitor always force full erasure, while complaint and rating evidence stays protected from inbox deletion.
55 built-in recipes, plus your own.
Complete light and dark color systems, glass controls, launcher shapes, uploaded art, sanitized inline SVG, and separate closed/open launcher icons.
🧰 A real operations inbox
Not a bot toy. A full agent inbox.
Whispers & AI draft
Invisible agent-to-agent notes the customer never sees, plus AI that drafts the reply for you to rewrite in your own voice.
Rich message cards
Up to 16 blocks: hero, buttons, fact lists, stats, tips, callouts, quotes, images and code. The AI can compose them for any saved reply.
Full agent toolkit
Emoji picker & reactions, voice input, attachments, typing indicator, snooze, archive, edit/delete and unread badges.
⚖️ How it compares
OpsIQ AI Chat vs a basic widget.
A fair capability matrix: what a premium, identity-aware AI chat gives you that a plain chat box never will.
| Capability | Basic chat widget | OpsIQ AI Chat |
|---|---|---|
| Agentic reasoning that plans each turn | — | ✓ |
| Live account lookups (real customer data) | — | ✓ |
| Safe AI actions (ticket / refund / route) | — | ✓ |
| Ticket written in the customer's own voice | — | ✓ |
| Capability controls for each ability | — | ✓ |
| Never fabricates (read-only, grounded) | Varies | ✓ |
| KB-grounded AI answers | — | ✓ |
| Identity-aware visitor card | — | ✓ |
| KB gap capture → drafted article | — | ✓ |
| Node-based flow builder | Limited | ✓ |
| Proactive rules + A/B testing | Limited | ✓ |
| Handoff carries full context → ticket | Email forward | ✓ |
| Rich message cards (16 blocks) | Plain text | ✓ |
| Multi-site from one workspace | Per-seat cost | ✓ |
| True white-label | Limited | ✓ |
| Asynchronous, non-blocking embed | Varies | ✓ |
| Automatic two-way translation | Limited | ✓ |
| Omnichannel gateway + one inbox | Separate inboxes | ✓ |
| AI continuity during human handoff | — | ✓ |
| Private by default (no trackers) | Varies | ✓ |
Full feature list
Everything in Client Chat AI.
Every capability, grouped. ★ marks a stand-out.
| Feature | What it does |
|---|---|
| Customer context | |
| Identity-Aware Visitor Card ★ | Customer identity arrives with a signed token: name, account status, last order and live journey, with no "verify your email" loop. |
| Order & History Context | Prior orders and chat history stay in view, so context never resets. |
| Agentic AI & live data | |
| Agentic Reasoning ★ NEW | Plans each turn: it decides what to check, then reasons before replying instead of pattern-matching a canned answer. |
| Live Account Lookups ★ NEW | Read-only tools pull the customer's real balance, invoices, services, domains, subscriptions and their own tickets. Everything is identity-scoped, only ever their own. |
| Safe AI Actions NEW | Opens a routed ticket in the customer's voice, files a refund for human approval, or hands off. Every action is permission-checked and capability-gated. |
| Capability Controls NEW | Turn each AI ability on or off per workspace (e.g. AI Ticket Awareness) through a connector-neutral capability contract, so it works the same across every connector. |
| Rolling Conversation Memory NEW | Long chats stay coherent because older turns are summarised, so context never falls off the end. |
| Anti-Fabrication Guardrails ★ | Read-only by default. It never claims an action it didn't take or quotes a number that isn't real. |
| AI & knowledge | |
| KB-Grounded AI Answers ★ | Grounded, not guessing. The AI stays within your documented facts and refuses to invent. |
| KB Gap Capture & Auto-Suggestions ★ | Every missed question becomes a drafted KB article ready to publish in one click. |
| KB Deflection Rate Tracking | Watch your AI-only resolution rate climb as you publish KB articles. |
| Automatic Language Detection NEW | Detects language from the visitor's message, with browser locale used only as an initial hint. Choose follow-customer, English-only, or an approved-language list. |
| Two-Way Live Translation ★ NEW | An operator can read an Arabic customer in English and type the reply in English; the customer receives Arabic automatically. Each side stays in their own language, originals remain preserved, and translations are cached. |
| Translated Widget & Automated Messages | Optionally translate the widget interface, handoff notices, closed-hours lines and other system messages. Post-chat email surveys can use the language of the conversation. |
| Escalation & handoff | |
| Department Routing | Escalations route to the right team: billing to billing, support to support. |
| AI Wait Companion ★ | The AI stays available while a visitor waits, knows elapsed wait and acknowledgement count, and cannot run write actions during the handoff. |
| Safe Timeout Return | Widget timer, boot polling and cron can return an unanswered or declined handoff to AI with a contextual continuation. |
| Human Follow-Up Policy | Configure 1–10 progressively spaced nudges, presence and business-hour guards, acknowledgement skipping and a maximum sequence window. |
| Working Hours & Offline Form | Set the team timezone and schedule, closed-hours copy, repeat timing, and a configurable offline form that can replace live chat while the team is away. |
| Automated Message Pools | Customize the name and email prompts, handoff and timeout notices, ticket lines, closed-hours wording, surveys and channel auto-replies; each moment can rotate approved variants. |
| Live Agent Inbox (PWA) | Agents run the inbox as a PWA, with full context, light/dark and push notifications. |
| Create Ticket from Chat | Chat escalates to a ticket with transcript, context and SLA timer attached. |
| Omnichannel inbox | |
| Seven Messaging Routes ★ | WhatsApp, SMS, Messenger, Instagram, Telegram, X and LINE enter the same Client Chat thread model alongside website chat. |
| Channel Identity & Dedupe | Preserves source, external identity, attachments and reply lineage while de-duplicating inbound events. |
| Per-Channel AI Policy | Enable AI replies independently by channel, keep account actions read-only or explicitly allow them, and debounce burst messages into one coherent response. |
| Network Reply Windows | Enforces WhatsApp and Meta reply windows, including WhatsApp re-engagement templates and the Messenger/Instagram HUMAN_AGENT extension for real teammates. |
| Proactive engagement | |
| Behavioral Trigger Engine ★ | Reach visitors at the right moment, based on URL, time, scroll, returning visitor and more. |
| Proactive Rules: 7 Action Types NEW | Chat, banner, tour, webhook, email, department assign and push: seven actions in one rule. |
| Proactive Rule A/B Testing NEW | A/B test rule variants and track reply rate live. |
| Automation & flows | |
| Chatbot Flow Builder (Node-Based) NEW | Node-based flow builder for deterministic branching conversations, with no code and a live preview. |
| Flow Trigger Types NEW | Seven triggers: greeting, keyword, after details, delay, on chat end, proactive rule and manual. |
| Flow Runtime Node Types (13) NEW | 13 runtime nodes cover messages, questions, branches, conditions, collection, knowledge, AI, email, tickets, delays and escalation. |
| Flow Variables & Interpolation NEW | Dynamic variables like {{visitor_name}} and {{page_title}} personalize flows. |
| Agent tools & management | |
| Saved Replies (Manual & AI) | Save replies manually or let AI generate rich cards. |
| Rich Message Cards NEW | Up to 16 blocks: hero, buttons, fact lists, stats, tips, quotes, images and code. |
| AI Draft & Rewrite NEW | AI drafts the reply; agents rewrite for tone, detail or length. |
| Typing Indicator NEW | Shows "typing…" while an agent composes, so it feels human. |
| Message Attachments & Upload Toggle NEW | Agents send files; per-thread control over client uploads. |
| Emoji Picker NEW | Visual emoji picker in widget and agent toolbar. |
| Emoji Reactions NEW | React to a message with an emoji for a quick acknowledgment without typing. |
| Whispers (Internal Notes) NEW | Invisible agent-to-agent thread notes the customer never sees. |
| Whisper Rooms | General and private admin rooms, mentions, room membership controls and AI-assisted internal collaboration stay beside the live inbox. |
| Inbox Workflow Controls | Configure first-response SLA, saved queue views, keyboard shortcuts, command palette, in-conversation search and customer-ticket history in the inspector. |
| Voice Input (Mic Button) NEW | Speak your message with voice-to-text in the widget. |
| Edit / Delete Message NEW | Fix typos or remove sent text (admin-gated). |
| Thread Snooze NEW | Hide threads until a set time; they auto-resurface. |
| Thread Archive NEW | Archive and unarchive completed chats. |
| Unread Message Badges NEW | Unread counts for admin and client; never miss a new message. |
| Chat Export NEW | Export transcript and attachments for compliance or records. |
| Quality & analytics | |
| CSAT / Post-Chat Rating NEW | Configurable post-chat rating prompt to measure CSAT and improve agent quality. |
| Answer-Level Feedback | Visitors can rate individual AI answers with thumbs up or down; feedback remains protected from ordinary inbox deletion. |
| Post-Chat Survey & Transcript | Control close timing, caps and questions, then let visitors email or download the conversation. Email surveys can follow the chat language. |
| Help Center Inside Chat | Let visitors search categories and published articles without leaving the widget, control entry placement and labels, and open articles on the source site, OpsIQ Help Center or your own Help Center domain. |
| Knowledge Training Controls | Crawl websites with page, depth and exclusion limits; configure local retrieval and embeddings; import knowledge; and maintain weighted, approved FAQs. |
| Deployment & customization | |
| Multi-Site Support ★ | One workspace for every property, each themed independently and all sharing one inbox. |
| Cross-Site Conversation Continuity | Visitors can continue the same conversation across connected properties when the workspace policy allows it. |
| White-Label Branding | True white-label: your colours, fonts and agent name. Remove "Powered by OpsIQ" on Business. |
| Light & Dark Theme Auto-Follow | Auto-follow the visitor's OS theme, or lock to a choice. |
| Async Widget Delivery ★ | Loads after the page becomes interactive and does not block the initial render. Polling remains available as the dependable realtime fallback. |
| PWA Inbox Installation | Install the agent inbox as a PWA with push. |
| Custom Brand Colors & Gradients NEW | Skinnable widget, buttons and inbox, with custom colours and gradients everywhere. |
| Custom Font & Copy NEW | Override the agent name, messages and labels per site. Every word can be yours. |
| 55 Premium Theme Recipes | Apply a complete light/dark color system, then customize glass blur, opacity, panel, bubbles, controls, chips, composer, launcher art and avatars. |
| Widget Position Presets NEW | Four corners or custom, fully responsive. |
| Security & privacy | |
| Private by Default ★ | No third-party trackers; identity via signed token; consent and retention enforced. |
| Registered-Origin Boundary | Only registered sites, the managed allowlist and the server's own origin receive cross-origin access. Copying the embed does not copy authorization. |
| Retention & Erasure Modes | Choose keep record, keep person or erase; scheduled retention and Forget Visitor force full erasure while complaint evidence remains accountable. |
| Realtime & Background Push | Choose polling, Pusher, Ably or Firebase for live delivery, plus Pusher Beams, OneSignal or VAPID for background admin alerts. |
| Current-Message Image Understanding ★ NEW | Opt-in vision inspects up to three JPEG, PNG, GIF or WebP images on the visitor's current message, up to 4 MB each. Files come from private attachment storage, past images are not resent, a vision-capable model is required, and failure falls back to a text-only reply. |
| Attachment Policy & Preview | Control upload size, attempts, windows and allowed extensions; enable previews, paste-only mode and per-thread client upload access independently. |
FAQ
Chat questions, answered.
Everything teams ask before they drop the widget on: performance, identity, handoff and customisation.
No. The widget loads asynchronously after the page is interactive and does not block the initial page render. Realtime delivery can use Pusher, Ably or Firebase, with polling as a dependable fallback.
No. OpsIQ detects language from the visitor’s message; browser locale is only an initial hint. An English-speaking operator sees an Arabic customer in English and writes back in English; the customer receives Arabic automatically. Each side stays in their own language, originals remain preserved, and translations are cached.
Logged-in customers are recognised via a signed identity token passed with the chat, the same mechanism behind logged-in-only targeting. Their name, account status, recent orders and journey show up on the visitor card, with no "verify your email" loop.
It says so and offers handoff. Configurable: route to a department, queue for an agent, or create a ticket with the conversation attached. Each missed question is also captured as a Pending KB suggestion: a drafted article you Approve & Publish in one click.
No. This is the customer-facing chat on your website. The Admin Chat AI is a separate widget your team uses in the backend to operate the business and engineer your servers by chatting: ask live billing, tickets and CRM and run signed, approve-queued actions, or connect a server and have it SSH in to read and edit code, run commands, monitor health and scan security, all backed up and audited.
Yes. You set it under AI Instructions in your OpsIQ admin, per surface (admin, customer, writing), with a live preview that shows the exact rendered prompt the AI will receive.
Yes. For a verified, signed-in customer it runs read-only lookups against their own account: balance, invoices, services, domains, subscriptions and their tickets. It's identity-scoped, so it only ever sees that customer's data, never anyone else's, and it never guesses when it can't confirm.
Yes, when you enable the opt-in vision setting and select a vision-capable model. It can inspect up to three JPEG, PNG, GIF or WebP images attached to the visitor's current message, then combine what it sees with approved knowledge and permitted account context. Each image is limited to 4 MB, files stay in private attachment storage, past images are not resent, and vision failure falls back to a text-only reply.
It can act, safely. When a customer asks, it opens a routed ticket written in their own first-person voice, files a refund request for a human to approve, or hands off to the right team. Anything that moves money or changes an account is filed for approval, never executed by the AI. And every ability is a capability you can turn on or off per workspace.
It's grounded and read-only by default. Answers come from your knowledge base plus the customer's own live data, not the open web, and hard guardrails stop it claiming an action it didn't take or quoting a ticket number or balance that isn't real. If it can't confirm something, it says so and offers a handoff.
Yes. One workspace runs every property you own. Each site gets its own theme, copy and proactive rules but shares a single agent inbox, knowledge base and analytics view. New sites inherit your base config automatically.
Only when something genuinely needs follow-up. The chat escalates into a native ticket with the full transcript, visitor card, journey and attachments carried over, and SLA timers start automatically. No re-explaining.
Yes. Your colours, gradients, fonts, copy, agent name and avatar are configurable from your OpsIQ admin. On Business plans you can remove "Powered by OpsIQ" entirely, and the widget follows light/dark automatically.