Products / AI Products

Answers from your documents, with the receipt attached.

AskDocs is a retrieval assistant over your own material. Every answer cites the paragraph it came from, and when the documents do not contain the answer, it says so instead of inventing one.

Built for

AI knowledge assistant for your own documents

  • Support teams answering the same policy question all day
  • Operations teams with manuals nobody can navigate
  • Anyone whose expertise is trapped in a shared drive

The problem

What this replaces.

The answer is in a PDF somewhere, and only two people know which one.
New starters ask the same fifteen questions for their first month.
We tried a chatbot and it confidently made things up.
Search returns forty documents and none of them are the one.

What's inside

4 modules that carry the work.

Grounded answers

Every claim traces back to a paragraph you can open.

  • Each answer cites its sources inline, linked to the exact section.
  • When retrieval finds nothing relevant, it says so rather than filling the gap.
  • Confidence is expressed by what it cites, not by a number nobody can interpret.
  • Answers can be pinned and corrected by a human, and the correction wins next time.

Your material

PDFs, docs, spreadsheets, wikis and web pages — indexed as they change.

  • Ingests PDF, Word, Excel, Markdown, HTML and plain text.
  • Tables and headings preserved, so a policy table does not become word soup.
  • Re-indexes on change, so a superseded policy stops being quoted.
  • Version-aware — it can tell you what the answer was before the change.

Permissions

People get answers from the documents they are allowed to read.

  • Access enforced at retrieval, not filtered out of the answer afterwards.
  • Document-level and folder-level permissions mapped from your existing groups.
  • Full query log — who asked what, and which sources answered it.
  • Nothing is used to train a third-party model.

Where people already are

A question asked in Slack should not require opening a new tab.

  • Web console, plus Slack and Teams where your team already talks.
  • Embeddable widget for an internal portal or a customer help centre.
  • API for putting grounded answers inside your own product.
  • Gap report — the questions people ask that your documents cannot answer.

What changes

The difference on the ground.

Cited

every answer links to its source paragraph

Says "I don't know"

when the documents do not cover it

Gap report

the questions your documentation is missing

Built with

Claude (Anthropic)Vector searchNode.jsMySQLSlack & Teams apps

You get the source. No licence to renew, no vendor who can switch it off.

Questions we get about AskDocs.

How do you stop it making things up?

It only answers from retrieved passages and cites them. If retrieval returns nothing relevant it refuses rather than generating. It is a real constraint, not a prompt asking nicely — but you should still spot-check answers on anything consequential.

Is our data used to train a model?

No. Your documents are indexed into your own store and passed as context at query time. They are not used for training by us or by the model provider.

What does it cost to run?

Per-query model cost, which depends on document size and traffic. We show you the projected monthly cost against your actual volume before you commit.

Can it handle scanned documents?

Yes, with OCR at ingest. Quality depends on the scan — we will run a sample of yours before making any promises about it.

Ready to move 20% faster?

Get your free custom strategy in 24 hours. No cost. No pressure. Just a clear plan.