AskDocsSelf-hosted memory for your documents.
Ask, instead of remembering where you filed it.
In production, running our own company. We can screen-share it today.
See it step by step

The problem
Documents are everywhere. Knowledge is not.
Answers sit scattered across folders and formats, they're hard to retrieve, they're outside any AI model's reach — and much of it can't leave the building. An answer you cannot trace is a rumour.
What we provide
Upload. Index. Ask. Verify. Sources arrive before the answer.
Many formats
PDF (with per-page OCR), Word, PowerPoint, Excel, Markdown, HTML, images, TIFF scans, plain text and code.
Page-precise citations
Every answer points to the exact page, highlighted — and says “not found” rather than guessing.
Hybrid search
Meaning-based and keyword search combined, then re-ranked, so the right passage comes first.
Workspaces and access control
Roles, per-document permissions, single sign-on and an append-only audit log.
Works with your AI tools
Built-in MCP server, so assistants like Claude, Cursor and VS Code can search your archive.
Self-hosted
Runs with Docker Compose on your own server; can run air-gapped.
Built on open source
PostgreSQL + pgvector
FastAPI
React
Ollama
Docker
What it replaces
Retrieval you can read back.
Instead of remembering where something was filed — or pasting documents into an outside AI tool — your team asks, and checks the cited page.
- Instead ofDigging through foldersYou runAsking a question
- Instead ofUploading files to outside AI toolsYou runAskDocs on your servers
- Instead ofAnswers you can't verifyYou runPage-precise citations
What's included
Set up for your team and looked after.
- Deployed on your own server with Docker Compose
- Your documents ingested, OCR'd and indexed
- Workspaces, roles and sign-on configured
- Sized for teams of up to about 50 people and 10,000 documents
- Run, patched and monitored by us
Status
LiveIn production, running our own company. We can screen-share it today.
