PaperQA2

What it does

PaperQA2 takes a set of local documents, including PDFs, text files, Microsoft Office files, and source code, and answers questions from them. It indexes and chunks the files, retrieves candidate passages, and produces a grounded answer with in-text citations. Metadata such as citation counts and journal quality is gathered from services including Semantic Scholar and Crossref.

People use it from the command line with the pqa tool or from Python code. Its default workflow lets a language agent search for papers, gather and rank evidence through re-ranking and contextual summarization, and generate an answer. Prior answers are indexed so earlier queries can be searched later.

The package suits researchers and developers who work with large collections of scientific literature and need cited answers. It is installed from PyPI and requires Python 3.11 or later. It calls an LLM, so users need an API key for a LiteLLM-compatible provider or a local open-source model server. Key limitations include model and API costs, public rate limits on metadata services without API keys, and the need to rebuild indexes created by older versions.

What it can do

• Answers questions over local PDF, text, Microsoft Office, and source code files with in-text citations
• Retrieves candidate papers via LLM-generated keyword queries and full-text search
• Ranks and re-scores evidence with LLM-based re-ranking and contextual summarization
• Fetches and stores paper metadata, including citation and journal quality data, from multiple providers
• Runs agentic workflows where a language agent iteratively refines queries and answers
• Provides an optional local embedding path via Sentence Transformers

Works with

• Semantic Scholar
• Crossref
• Unpaywall
• OpenAlex
• LiteLLM

Before you choose

• Pricing depends on the LLM and metadata API providers you use; the code itself carries no listed fee.
• Docs objects pickled from versions before version 5 are incompatible and must be rebuilt.
• Version naming changed in December 2025 to calendar versioning, while the PaperQA2 name still refers to paper-qa 5 and later.

Listed under

• Data
• Research

Get started

Requires an LLM API key or a local open-source LLM server.

1. Install the package with pip install paper-qa>=5 in a Python 3.11+ environment.
2. Set an LLM API key environment variable (for example OPENAI_API_KEY) or configure an open-source LLM server.
3. Optionally set CROSSREF_API_KEY and SEMANTIC_SCHOLAR_API_KEY for large indexing jobs to avoid public rate limits.

The code is open source under Apache-2.0. Costs come from the LLM and search/metadata APIs used at runtime, which are billed by those providers.

PaperQA2 is free to use. See their pricing page ↗

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• GitHub - Future-House/paper-qa: High accuracy RAG for answering questions from scientific documents with citations · GitHub ↗github.com
• PaperQA2 achieves SOTA performance on RAG-QA Arena | FutureHouse ↗www.futurehouse.org
• FutureHouse ↗www.futurehouse.org
• What is an AI Scientist? | FutureHouse ↗www.futurehouse.org

Company website

github.com: product, pricing and policy pages

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

manual_review

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.