AI grounded in YOUR knowledge

An AI chatbot that knows your product cold.

Drop in your help articles, policies, and FAQs. The bot answers visitor questions using your actual content — not made-up guesses — and hands off to a human the moment things get specific.

Free starter replies included Bring your own OpenAI key 12+ languages
🤖

Acme AI

Always available · trained on 247 articles

Do you offer refunds for annual plans?
Yes — full refund within the first 30 days, prorated after that. The 30-day window starts on the original charge date, not the renewal.
Source: Refund policy · KB article
Can I get a custom enterprise SLA?
  Connecting you to Sarah from Sales…

AI that answers, not deflects

62%tickets auto-resolved
<2smedian answer time
12+supported languages
24/7coverage, no shifts
Why a grounded AI

Generic LLM bots make things up. Yours won't.

The AI only answers from articles you've published. If your KB doesn't cover it, the bot says so and routes the visitor to a human — instead of confidently inventing a wrong answer.

Grounded in your KB

RAG retrieval pulls relevant article snippets per question. Answers cite which article they came from, so visitors trust them.

Smart human handoff

When the AI's confidence drops or the visitor asks for a human, it transitions cleanly with full conversation context.

Predictable cost

Per-site daily cap on AI replies, plus an option to bring your own OpenAI key (no markup) once you scale.

How it works

Real AI, configured in minutes.

No vector-database setup. No retrieval pipeline to debug. No prompt engineering.

Knowledge base training

Train it the same way you'd train a new hire.

Drop articles into the Knowledge Base — markdown, FAQ pairs, scraped URLs, or pasted text. The AI re-indexes immediately. Every answer cites the article it pulled from, so admins can see exactly which content is doing the work.

  • Markdown editor with live preview, status workflow (draft / visible / hidden)
  • Bulk import — paste multiple docs separated by ---
  • One-click migration from existing AI training sources
  • Per-article toggle: "Use for AI" + "Visible in public help center" are independent
  • Search analytics show which questions your KB doesn't cover yet
  Knowledge Base
Refund policy visible · 1.2k views
Shipping & delivery visible · 980 views
Enterprise SLA draft
Account API tokens visible · 412 views
Smart handoff

The bot knows when to step aside.

Visitors don't get trapped in a bot loop. When the AI hits its confidence threshold or the visitor asks for a human, the conversation hands off mid-flight with full transcript context attached. Your agents pick up exactly where the AI left off.

  • Auto-handoff on detected confusion or repeated visitor frustration
  • Explicit handoff on phrases like "talk to a human" / "real person"
  • Agent sees full bot transcript + handoff reason before accepting
  • Handoff log per session for QA review
  • Out-of-hours behavior: bot collects email + sends offline message
  Handoff log
10:42 · Bot answered 3 questions
10:43 · Visitor: "I need a real person"
10:43 · Handed off to Sarah
10:44 · Sarah: "Hi! I see you were asking about enterprise SLA..."
Multilingual

Answers in the visitor's language — even if your KB isn't.

Visitor types in French, the AI answers in French, even if your articles are only in English. The translation happens in-flight, so you don't need to maintain N copies of every article. Agents see both the original and translated text in the console.

  • 12+ languages out of the box (EN / ES / FR / DE / PT / IT / NL / PL / JA / KO / ZH / AR)
  • Auto-detect from inbound message OR honor visitor's browser locale
  • Agents see original + translated side by side
  • Translate any agent-typed reply with one click before sending
  Multilingual
Bonjour, comment puis-je vous aider ?
Avez-vous une politique de remboursement ?
Oui — remboursement intégral sous 30 jours.
Source article (EN): "Refund policy"
Cost control

Two pricing models. Your call.

Start with our managed plan: we provide the OpenAI key, you pay per-reply bundles with a daily cap so a runaway bot never blows your budget. Outgrown that? Switch to bring-your-own-key — you pay OpenAI directly at their cost, we charge zero markup.

  • Managed: Reply packs from 200/month. Per-site daily cap.
  • Own-key: Plug your OpenAI key, no markup, unlimited replies.
  • Plan gate prevents the bot using a paid model on a free-plan site (no surprise bills)
  • Usage analytics: replies, tokens, est. cost per day
  This month
Replies served   142 / 500
Tickets deflected   88
Avg cost / reply   $0.018
Daily cap   25 left
3 steps to live AI

From signup to first AI reply in 10 minutes.

No infrastructure. No prompt tuning. No vector store to maintain.

Train it

Paste FAQs or upload markdown articles into your Knowledge Base. The AI indexes immediately — no embedding job to wait on.

Configure it

Set the bot persona ("friendly", "formal"), choose languages, set the daily reply cap. Pick managed AI or bring-your-own-key.

Deploy it

Already shown in your widget. Toggle "AI on" and the bot starts answering. Watch the deflection rate climb in your dashboard.

Why this AI

Not all AI chatbots are equal.

Built-in is different from "we wrapped GPT and called it support."

  MyLiveChat AI Generic LLM bot FAQ-only bot
Grounded in your articles (RAG)Limited
Cites source article in answer
Smart handoff to live agentOften dead-end
Multilingual (12+ languages)Per-language config
Per-site cost capPay-per-tokenFlat fee
Bring your own OpenAI key
Learns from past chat transcriptsGeneric context
Handoff log + analytics
"Our refund-policy questions used to consume 30% of our team's time. With MyLiveChat AI grounded in our actual policy docs, the bot now handles 90% of them on its own — with citations, in seconds. Our team gets to focus on the questions that actually need them."
PR
Priya R. Director of Support, e-commerce SaaS
Common questions

Frequently asked.

How accurate are the answers?
Accuracy is bounded by your KB. Because the AI cites the source article it pulled from, you can audit accuracy directly — if an answer is wrong, click through to the source article and fix that one place. Most customers see >90% accuracy on policy / pricing / how-to questions within a week of publishing 20-30 articles.
What about my data — is it sent to OpenAI?
On the managed plan: yes, conversation snippets + relevant KB chunks go to OpenAI to generate the response. We send the minimum needed (no historical transcripts unless explicitly enabled). On the bring-your-own-key plan you own the OpenAI relationship directly — bound by their data-use policy, which excludes API content from training by default.
What happens when the bot doesn't know the answer?
Two paths. Inside business hours: the bot says "I'm not sure — let me get someone who can help" and hands off to a live agent with the full conversation context. Out of hours: it offers to email a human for the visitor and collects their email + question.
How much does the AI cost?
Free tier includes a starter pack of replies each month. Paid reply packs start at $9 / month for 200 replies. Or bring your own OpenAI key and pay OpenAI directly at their list rate — we charge zero markup. Most customers find their effective cost is <$0.02 per resolved conversation.
Can I see what the bot has been answering?
Yes — the AI Chat Log dashboard shows every bot conversation, the article it cited, the visitor's reaction, and any handoff that happened. Filter by date, by article, or by handoff reason for QA review.
Can I disable the AI for some pages or visitors?
Yes. Per-page rules, per-visitor segment overrides, and operating-hours based on/off. Some customers run AI only on the public site and humans-only for authenticated app users — takes one toggle per visitor segment.

Let your AI do the routine work.

Free starter replies. 10-minute setup. Zero infrastructure.

Chat with your visitors, Increase sales and conversions, make your customers happy!