Paper2Agent is an Open Source framework developed by researchers at Stanford University (Jiacheng Miao, James Zou et al.) that transforms scientific papers and their associated code into interactive AI agents. Multiple specialized agents analyze the paper and codebase to automatically generate a Model-Context-Protocol (MCP) server with tested tools, resources, and prompts that can connect with coding agents like Claude Code or Codex. The method was published in Nature on September 16, 2026, and validated on a 100-paper biology suite, with 74 papers successfully agentified and the generated agents achieving 91.2% accuracy on 300 Benchmark questions. The code is freely available on GitHub under MIT license, and additionally there is a

Belege (2)