Website crawl + file upload
Help-center pages, API reference, and PDFs in one knowledge base. Answers stay tenant-scoped.
For SaaS & docs
An AI chatbot trained on your help center — not on the open internet
Product docs exist. Support still repeats the same setup, billing, and API questions. Point us at your public docs and PDFs; the bot answers from that corpus and hands off when the article does not cover the case.
- RAG over your docs
- Exact-match FAQs first
- 7-day money-back on paid
Why docs do not deflect
Search in the help center is not the same as answering in the product
Users do not browse three levels of navigation when they are blocked on an API key or a failed webhook. They type a question into chat — and generic Copilot-style answers invent version numbers you never shipped.
A grounded bot is a retrieval layer over *your* articles, changelogs, and FAQs. Wrong or missing docs show up as missed retrieval, which is fixable. Hallucinated product behaviour is not.
- Hybrid search catches SKUs, error codes, and exact endpoint names that pure semantic search misses
- FAQ exact-match answers known issues without an LLM round-trip
- Out-of-scope questions decline instead of guessing a workaround
How it works
From help-center URL to in-product answers
Crawl the docs site and upload the rest
Point at your public docs URL. Add PDFs, Markdown, and CSVs for runbooks that are not on the site. Re-crawl when you ship.
Pin exact FAQs for the expensive tickets
Billing, SSO, and known defects get an exact-match FAQ so the model cannot paraphrase them into the wrong plan name.
Embed in-app and keep humans in the loop
Paste the widget on docs or inside the app. Agents take over with the thread when the answer needs an account-specific exception.
SaaS features
Retrieval that respects how technical content actually looks
Hybrid keyword + semantic search
Error codes and function names match even when the user's wording does not match the article title.
Reranking before the reply
A second pass reorders retrieved chunks so the model sees the most relevant passage first.
Handoff with notes
When docs are silent, the conversation goes to your team with history — not a dead-end “I don't know” with no ticket.
Channels your users already use
Website widget, plus WhatsApp and Instagram on Growth and above. Same knowledge base across channels.
Analytics on what docs miss
See which questions fail retrieval so you fix the article once instead of answering it forever.
Compare
Docs chatbot vs generic assistant vs in-house RAG
| Capability | Reply | Generic LLM widget | Build RAG yourself |
|---|---|---|---|
| Answers from your corpus only | YesYes | No | If you scope it |
| Declines when docs don't cover it | YesYes | No | If you prompt it |
| Hybrid search for codes & names | YesYes | Rare | You implement BM25 |
| Time to first useful answer | ~15 minutes | Same day, higher risk | Months |
Answers from your corpus only
- Reply
- YesYes
- Generic LLM widget
- No
- Build RAG yourself
- If you scope it
Declines when docs don't cover it
- Reply
- YesYes
- Generic LLM widget
- No
- Build RAG yourself
- If you prompt it
Hybrid search for codes & names
- Reply
- YesYes
- Generic LLM widget
- Rare
- Build RAG yourself
- You implement BM25
Time to first useful answer
- Reply
- ~15 minutes
- Generic LLM widget
- Same day, higher risk
- Build RAG yourself
- Months
FAQ
SaaS docs chatbot FAQ
Today you crawl public URLs and upload files (PDF, DOCX, TXT, MD, CSV). If the page is behind login, export or upload it. We do not claim a live Notion connector on this page.
It should not. Retrieval is over your ingested content; uncovered questions are declined or handed off. Garbage docs still produce garbage answers — fix the source.
No. It sits on top of it. Search and articles still matter; the bot is for people who will not browse.
Growth is the usual landing plan when you want WhatsApp plus hybrid search. Compare limits on the pricing page.
API read is on Growth; write is on Business. Programmatic keys are plan-gated server-side.