PaperQA2
Status:
watch-list| Maturity:emerging| Last reviewed: 2026-03-04
What it does
PaperQA2 is an AI agent developed by FutureHouse (now managed by Edison Scientific) optimised for retrieving and summarising information from scientific literature. It uses an agentic approach to search, retrieve, and synthesise information across papers, providing grounded answers with citations.
Key capabilities
- Agentic literature retrieval and synthesis across scientific papers
- Source-grounded answers with inline citations
- Designed to minimise hallucination through retrieval-augmented generation
- Open source — can be self-hosted or integrated into research workflows
Fit for Sheffield researchers
High potential value for STEM and health researchers. The open-source nature enables institutional deployment with appropriate data governance controls. Requires technical expertise to deploy and configure — not suitable as a self-service tool for general researchers without IT/research computing support.
Licensing and access
Open source model available via GitHub. Hosted service options may exist through Edison Scientific. Recommend exploration by research computing or IT Services for potential institutional deployment.
Trust and governance notes
Open-source deployment enables on-premise or cloud-with-governance hosting, which may address GDPR concerns better than proprietary alternatives. Data handling depends on deployment configuration.
Observed limitations
- High technical barrier to deployment and use
- Primarily STEM-focused corpus
- Actively developing — interface and capabilities may change significantly
- Requires researcher curation of input papers for best results
Related tools
Sources and evidence
- FutureHouse PaperQA2 announcement: https://www.futurehouse.org/research-announcements/wikicrow (accessed 2026-03-04)