Stanford turns 100 research papers into AI agents with Paper2Agent

A Stanford team published Paper2Agent in Nature, an automated pipeline that converts research papers into MCP agents that rerun the paper's own methods on demand. Of 100 computational biology papers, 74 became working agents with 593 of 599 validated tools, beating handing the same repos to Claude (91.2% vs 82.7%).

• 74 of 100 papers converted into agents; 599 tools extracted, 593 validated
• Agents scored 91.2% vs 82.7% for Claude with the same repos on 300 questions
• AlphaGenome agent: 22 tools in ~45 minutes for ~$14, 100% on 15 novel queries
• Three agents from unrelated papers jointly flagged GPR137 as a likely psoriasis gene