How it works
From a URL to answers, and kept current
Start with a URL
It finds your sitemap, ranks the pages that matter — pricing before careers — and reads them.
- No FAQ spreadsheet to fill in
- Nothing to re-type
yourbusiness.comLive- Reading your siteSitemap found · 38 pages ranked
- Learning your businessServices, prices, hours
- Drafting your assistantName, tone, greeting, questions
- Rehearsing before it's liveTest questions answered
- Ready to pasteOne line of HTML
Add what isn't on the site
Price lists, menus, policies: upload PDF, Word, text, Markdown or CSV and watch each one index.
- Status for every source
- A failed file tells you why
Answers trace back to a passage
Each reply is built from the passages that match the question, so you can see exactly where it came from.
- Fix a page, fix the answer
- No prompt-tuning
Searching 42 pages for “refund after 30 days”
/returns
Refunds run for 14 days from the date of purchase.
/faq
After 14 days we review each case individually.
It notices when pages change
Pages are re-checked on a schedule; unchanged ones cost nothing, changed ones are re-read.
- Weekly by default
- Polite to your server
- Reading your siteSitemap found · 38 pages ranked
- Learning your businessServices, prices, hours
- Drafting your assistantName, tone, greeting, questions
- Rehearsing before it's liveTest questions answered
- Ready to pasteOne line of HTML
Under the hood
The details that make it work
Sitemaps, followed
Sitemap indexes, compressed sitemaps and the sitemap named in robots.txt.
robots.txt respected
Pages you've disallowed are never fetched — and never count against your budget.
Polite crawling
One request at a time per site, with the delay your robots.txt asks for.
Structured data read
JSON-LD and canonical tags on your pages are used, not ignored.
Heading-aware passages
Content is split along your headings, so an answer keeps its context.
Originals kept
Every document you upload can be downloaded again, exactly as you sent it.
The rest of the product
It works as one
AI agents
Grounded answers with the source attached, a clean handover to your team, and no confident guesses.
Explore ai agentsAutomations
Bookings on your real calendar, leads where your team works, and calls to your own systems.
Explore automationsAnalytics
Every conversation read, scored and summed up, with this week's most useful fix at the top.
Explore analytics
Good questions
What people ask about knowledge before they start.
Still deciding?Ask the people who build it.What can it read?
Public web pages (from a URL or your sitemap) and uploaded PDF, Word, plain-text, Markdown and CSV files.Does it work on sites built with React or Vue?
It reads the HTML your server sends. A site that only renders its content in the browser with JavaScript may index as nearly empty — the dashboard flags that, and uploading the key pages as documents is the workaround.What about images and scanned PDFs?
Text in an image, or a PDF with no text layer, can't be read. The source is marked failed with that reason rather than silently contributing nothing.How often does it re-read my site?
Pages are re-checked weekly. The check asks your server whether the page changed, so an unchanged page costs almost nothing.
