Stanford team introduces Paper2Agent: AI tool that transforms research papers into functional AI agents

A Stanford research team, led by Jiacheng Miao and James Zou, has introduced Paper2Agent, a tool that converts academic papers and their associated codebases into executable AI agents. Published in Nature on September 16, 2026, the system enables agents like Claude Code to run the methods described in research papers through natural language commands. The tool, available under an MIT license, is deployable as a skill for Claude Code and Codex, with prebuilt servers hosted on Hugging Face Spaces. Paper2Agent operates through a six-step process managed by a central orchestrator, validating tools with strict criteria including file accuracy, numerical consistency, and perceptual hash matching. The system successfully validated 22 tools for AlphaGenome in under 45 minutes, achieving high performance on benchmark tasks. The tool also demonstrated robustness against various errors, including dependency and API failures, and was used to re-examine genetic variants, offering insights into causal gene assignment.