PaperQA is an open-source retrieval-augmented generation package from FutureHouse built for one job: answering questions from scientific documents with high accuracy and proper citations. Released under the Apache-2.0 license, it has earned more than 9,400 stars on GitHub and was actively developed through 2026.
Generic RAG pipelines tend to blur details and drop references, which is fatal for scientific work. PaperQA instead runs an agentic pipeline over your papers: it searches the corpus, gathers evidence snippets, scores their relevance, and composes an answer in which claims are tied to specific sources, mimicking how a researcher reads. The package (in its PaperQA2 generation) supports PDFs and other document formats, works with multiple LLM and embedding providers, and exposes both a Python API and command-line tools for building literature-grounded assistants. FutureHouse reports that PaperQA2 exceeds human-expert performance on several literature-review benchmarks, and the system can also pull in full-text metadata such as authors, journals and citation counts, which helps it judge which sources deserve the most weight in an answer.
PaperQA is free and open source under the Apache-2.0 license. You pay only for the LLM and embedding API calls you choose to make, or nothing when running local models.
Developers can also tune the evidence-retrieval and answer-generation steps independently, which makes PaperQA a common building block inside larger scientific copilots, literature-monitoring tools and automated evidence-review pipelines rather than only a standalone question-answering script.
It is for researchers, R&D teams and builders of scientific assistants who need answers they can verify โ every claim traceable back to the paper it came from.