Stanford's Paper2Agent turns a research paper into an AI agent

For roughly US$14 in compute and about 45 minutes of build time, a Stanford system called Paper2Agent turned the AlphaGenome paper into an AI agent that scientists could talk to in plain English.

Nature's Kaia Glickman reports that the tool, from a group led by Stanford computer scientist James Zou, "starts by accessing a paper's main text, code, data sets and other elements." Those materials are deposited onto a digital platform called an MCP server. A team of AI agents then autonomously builds tools that can apply the paper's methods to fresh data and places those tools on the same server. Scientists connect using a large language model of their choice and query the resulting paper-specific agent conversationally.

On the AlphaGenome test, the agent "passed its initial test with flying colours, answering genetics questions with near-perfect accuracy," Glickman writes. It also outscored Biomni, a tool developed by academic researchers that draws from dozens of databases and was given the same paper and the same questions.

One answer stood out. Asked to identify a gene explaining why a DNA change is linked to bad cholesterol, the agent named a different causal gene from the one pinpointed in the original AlphaGenome paper. The Nature piece frames this as the agent helping scientists reevaluate published conclusions rather than as a miss.

Zou said the technology "can help us to reimagine what knowledge looks like in the future."

Three of the AI researchers in our Who's Who tracker shared the Nature link the same week it ran.