Paper2Agent converts research papers into interactive AI agents

A Stanford team has built a system that takes a research paper and turns it into a working AI agent, one that can reproduce the paper's results, apply its methods to new data, and hold a conversation with agents made from other papers. The tool, called Paper2Agent, is described in a study published in Nature on 16 September 2026 and reported by Nature News.

The setup is prosaic. Paper2Agent "starts by accessing a paper's main text, code, data sets and other elements," according to the article, then deposits that material on what is called an MCP server. A team of AI agents autonomously writes tools that apply the paper's methods to fresh data, and scientists connect through a large language model of their choice.

The showcase is genetics. "Paper2Agent autonomously created an agent for the AlphaGenome paper in about 45 minutes, and the required computing power cost US$14," the Nature News piece reports. On the same set of questions, the resulting agent "bested the scores of other top biomedical AI agents that were given access to the same paper." That included Biomni, a rival built by academic researchers that "draws from dozens of databases but scored far lower than the AlphaGenome agent."

James Zou, a computer scientist at Stanford University and a co-author of the paper, told Nature News the technology "can help us to reimagine what knowledge looks like in the future." Applied to a DNA change linked to 'bad' cholesterol, the AlphaGenome agent nominated a different causal gene than the original paper did, a divergence Nature News notes is supported by data on either side.