Paperbrief

Upload PDFs, contracts, research notes. Ask grounded questions. Get answers with citations back to the source chunks — built on Claude + pgvector.

What Paperbrief is NOT

Not a document management system. No folder hierarchies, no permissioning beyond org boundaries, no versioning. Use it as a layer on top of wherever your docs actually live.
Not a search engine. It's question-answering — it gives you a synthesised answer with citations, not a ranked list of matches.
Not freeform LLM with your docs in the prompt. It uses retrieval + grounding to keep answers tied to source content. You can't make it write you a poem.

File size cap: 25 MB. We'll raise on request if there's a real use case.
OCR: capped at 50 pages per document (cost control). Beyond 50 pages, only the first 50 are searchable.
Chat history is org-scoped, not user-scoped — anyone in your organisation can see all conversations. Per-user privacy is on the roadmap.
Markdown rendering of answers is plain-text today (no bold, lists, code blocks). Improving in a follow-up.

If any of those is a deal-breaker, email us — we want to know what to prioritise.