Let visitors ask your documentation a question directly
Search boxes make people guess the right keyword. JestBot lets visitors just ask — it retrieves the relevant section of your docs and answers grounded in it, with the source chunk visible for you to verify.
No card required · Free plan included
< 2 min
Setup time
PDF · DOCX · MD · URL
Doc formats
Cosine similarity
Retrieval method
No card needed
Free plan
Why a search box isn't enough
Keyword search makes people do the translation work
A visitor with a real question has to guess the right keyword, scan several pages, and piece the answer together themselves — or give up and open a support ticket.
Keyword search misses phrased questions
"How do I reset a webhook after a failed delivery" doesn't match well against a page titled "Webhooks Configuration" — even though the answer's right there.
Long docs bury short answers
A one-paragraph answer to a specific question is often buried inside a much longer reference page a visitor has to read end to end.
Unanswered questions become support tickets
Every question your docs could have answered directly, but didn't, ends up as a ticket someone on your team has to personally resolve.
Live on your docs site in three steps
01
Ingest your documentation
Paste your docs URLs, upload Markdown or PDF exports, or paste raw text — JestBot chunks and embeds it automatically.
02
Embed on your docs site
One script tag before </body>, and the chat bubble appears alongside your existing search.
03
Verify answers in Test Chat
Check the retrieved chunks and similarity scores for real questions before rolling it out publicly.
Why it works for documentation
Retrieval built for reference content specifically
Docs are exactly the kind of structured, factual content retrieval-grounded AI is best at — precise questions, precise sourced answers.
Ingest docs in whatever format they're in
Paste your docs site URL, upload Markdown or PDF exports, or paste raw text. Every source is chunked into overlapping windows automatically, so long reference pages don't bury the specific answer a visitor needs.
- URL, file, or text ingestion
- Automatic chunking with overlap
- Works with existing docs, no rewrite needed
See the exact section it used
Every answer is built from the specific chunks retrieved for that question, visible with a similarity score in Test Chat — so you can verify accuracy against your actual docs before trusting it in production.
- Cosine-similarity retrieval
- Retrieved chunks visible for QA
- Streaming responses over SSE
Build a custom UI if you want one
The chat endpoint that powers the widget is a documented public API — authenticate with your bot's public key and stream responses over SSE into your own docs site's native chat UI instead of the default widget.
- Documented public chat API
- Public key safe to expose client-side
- Session IDs keep recent context automatically
Keep answers current as docs change
Re-ingest updated pages anytime from the dashboard, or delete outdated documents individually, so answers never lag behind a documentation update.
- Re-ingest anytime
- Per-document deletion
- No stale answers from old versions
Where teams put it to work
One bot, wherever your documentation lives
The same setup works for a product help center, an API reference, or internal engineering docs.
Product help centers
Let customers ask a question directly instead of scanning a help center's article list.
Developer & API references
Ground answers in your actual API docs so integration questions get precise, current answers.
Ticket deflection
Answer the questions your docs already cover so support only sees what's genuinely undocumented.
Onboarding guides
Walk new users through setup steps grounded in your real getting-started docs.
Multilingual docs support
Answer in the language a visitor writes in from one English (or any language) knowledge base.
Escalate what docs can't answer
Hand off to a human the moment a question needs judgment your documentation doesn't cover.
“Being able to see the retrieved chunks in Test Chat is what sold me — I could actually verify the bot was citing the right page before we put it live on our docs.”
Marcus T.
Developer relations
Questions people actually ask
JestBot — AI Chatbot for Documentation