An AI agent that answers from what you’ve already written.
It replies in the Messenger from your help center, articles, files and pages, in the tone you choose, and hands the conversation to your team when it should. In the inbox it drafts, rewrites, translates and summarizes.
Answers with your knowledge
Help center guides, articles, files, web pages and the answers your team already gave.
Knows when to step back
It hands off to a person when it isn’t sure, and only answers the people you allow: signed-in users, or visitors too.
A copilot for the team
Draft a reply, rewrite it, translate it or summarize a long thread, then send it yourself.
Tune it before it talks
Set its instructions and tone, then try it in the playground on real questions before any customer sees an answer.
- A warm, neutral or formal tone
- Your own instructions
- A playground to test answers
Find what it doesn’t know
Gaps lists the questions it couldn’t answer, so you know which guide to write next.
- Each unanswered question, listed
- Mark a gap answered or dismissed
- New help center guides join its sources
Let it look things up
AI actions call your own API, with GET or POST, so it can check an invoice or an account before it answers.
- Your own endpoints
- Secrets encrypted at rest
- Every run recorded with its cost
Included replies, then credits
Each plan includes AI replies every month. Past that, each reply uses one credit from your prepaid balance, and credits last 12 months.
See pricingQuestions
Talk to usA Claude model, reached through zero-data-retention routing, so the provider keeps neither prompts nor replies.
Only where you turn it on, for the people you choose. Elsewhere it drafts and your team sends.
Prompts, replies and knowledge never reach our logs or analytics. We keep ids, counts and the cost of each AI run.
Let the AI take the first answer.
Add your help center as a source, test it in the playground, then switch it on in the Messenger.