Stanford's Paper2Agent turns research papers into AI agents

A framework from Stanford converts a static research paper into a working AI agent in about 45 minutes for roughly $14, using the Model Context Protocol to wrap the paper's manuscript, code, data and workflows behind a server a chat agent like Claude Code can call. Paper2Agent, described in Nature on 16 September 2026 by Jiacheng Miao, Joe R. Davis, Yaohui Zhang, Jonathan K. Pritchard and James Zou, was applied to 100 computational biology papers; 74 converted successfully, producing 593 validated tools.

On the AlphaGenome case study the agent scored 98.7% on tutorial-derived queries against 82.7% for direct repository access, and across benchmark questions the system averaged 91.2% accuracy. The paper pitches each agent as a 'virtual corresponding author' that can answer questions about a paper, apply its methods to new data, and collaborate with agents built from other papers.

Five researchers we track posted the link in the days after publication.

Zou frames the shift bluntly. 'Knowledge should not be static records. It really should be dynamic and interactive,' he told IEEE Spectrum. Olivier Elemento of Weill Cornell Medicine called it 'a real advance in terms of how we think about the publication process.' Zou is also careful about credit: 'It's still important to attribute the final discoveries and reference them back to original papers and original human authors.'

The 26 conversion failures came from incomplete code, missing documentation or incompatible dependencies, effectively an automated audit of a paper's reproducibility. Neither the paper nor the coverage says whether source-paper authors are asked to consent before their work is agentified, or how well the method travels beyond computational biology.