Paper2Agent turns any research paper into a working AI agent

Paper2Agent, a Stanford system covered by Nature on 16 September, converts a research paper into an AI agent by loading its text, code and datasets onto an MCP server and turning a team of AI agents loose to build tools that apply the paper's methodology to fresh data. A scientist then connects their preferred large language model and queries the paper-specific agent in ordinary language.

'Can help us to reimagine what knowledge looks like in the future,' says James Zou, a computer scientist at Stanford University and a co-author of the work.

On the AlphaGenome paper, a predictor of DNA-sequence properties, the system built 22 tools in about 45 minutes for US$14. All 22 passed automated validation without human intervention. The resulting agent hit near-perfect accuracy on genetics questions and outscored Biomni, an established biomedical AI tool developed by academic researchers.

In one test the AlphaGenome agent identified a different causal gene than the original paper had, for a DNA change linked to cholesterol levels. That is a demonstration, in the Nature write-up's framing, that a paper's own conclusions can be re-examined without designing new experiments. Three researchers we follow in our Who's Who directory shared the piece the day it appeared.