PaperQA2

What it is for

PaperQA2 is an Apache-2.0 RAG library for scientific PDFs: install paper-qa, point it at a folder of papers, and ask a question with citations. Use it when the job is answer from papers you already have, in your own runtime, not when you want a hosted flashcard site.

What makes it unique: the official GitHub file list (cover) is the product surface — src/paperqa, Apache-2.0, CLI + library — not a marketing poster.

Key features

• CLI over a paper folder — pip install paper-qa, then ask from the directory (README).
• Metadata + retraction check — README quickstart pulls citation counts when it can.
• Full-text index — parse and cache PDFs (and text / Office / code files) for RAG.
• Agentic add/query — library API to add docs and ask with an LLM agent.
• Local or hosted models — you pass the model; local embeddings are documented.

Who it's for

Best fit: engineers and labs who will run Python and already have a PDF pile.

Pricing

Verified 2026-10-22 (Asia/Shanghai) from the Future-House/paper-qa README and LICENSE.

• Software — Apache-2.0. License fee $0.
• Hosted PaperQA seat — none on the public repo. Do not invent a cloud dollar.
• Model / embedding APIs — you bring keys; those vendors bill separately.

Limitations / costs

• You bring the model bill. Apache-2.0 does not include tokens.
• Library/CLI, not a hosted library UI. For a browser workspace, see Anara.
• Not a citation map. For neighborhoods, see Litmaps.

Alternatives

• Elicit — hosted research-agent composer over papers.
• Anara — hosted PDF folders and chat.
• STORM — MIT topic-to-article engine with a research preview.

Official tutorial

Written path: the README Quickstart (install, folder of PDFs, ask).

Website

https://github.com/Future-House/paper-qa

Source

Catalog row curated for AI Tools Nav. Cover is the official Future-House/paper-qa GitHub file list. License Apache-2.0; software $0. Tutorial: README Quickstart.

Last verified

2026-10-22 (Asia/Shanghai) editorial check.