Stanford's Paper2Agent turns any paper into a callable AI agent

A Stanford team has published a system that converts a research paper into an AI agent scientists can converse with, apply to new data, and connect to agents built from other papers. Writing in Nature, computer scientist James Zou and colleagues call the tool Paper2Agent.

The pipeline reads a paper's text, code and data and deposits them onto an MCP (Model Context Protocol) server, where a group of AI agents autonomously builds tools that apply the paper's methods to fresh data. A scientist then plugs in their preferred large language model and asks questions in natural language.

The showcase test was AlphaGenome. "Paper2Agent autonomously created an agent for the AlphaGenome paper in about 45 minutes, and the required computing power cost US$14," the article reports. "The agent passed its initial test with flying colours, answering genetics questions with near-perfect accuracy." It outscored Biomni, a tool built by academic researchers that draws from dozens of databases.

More interesting than the price: pointed at a cholesterol question, the AlphaGenome agent named a different gene as the culprit than the paper it was built from. The article frames the discrepancy as a feature, saying scientists "can use the tool to re-evaluate published conclusions without designing entirely new experiments."

Zou says the technology "can help us to reimagine what knowledge looks like in the future." Three researchers we track posted the piece the day it appeared.