Paper2Agent Turns Research Papers Into AI Agents

Researchers at Stanford University introduced Paper2Agent, a framework that converts research papers and their code into interactive AI agents, in a paper published in Nature. Users can ask the resulting agents to explain methods, reproduce analyses or apply workflows to new data through natural language.

Paper2Agent examines a manuscript, supplementary materials and its codebase, then creates a Model Context Protocol server containing executable tools, resources and workflow prompts. Specialized agents configure the software environment, extract tools and test outputs against the original results. Tools that repeatedly fail validation are excluded.

The system successfully converted 74 of 100 computational biology papers and validated 593 of 599 proposed tools. On 300 questions based on tutorials, Paper2Agent using Claude Sonnet 4 recorded 91.2 percent accuracy, compared with 80.3 percent for Claude Code using the same model with direct access to each paper and repository. Tests also covered AlphaGenome, Scanpy and research outside biology.

Some papers could not be converted because of missing code, data, model files or unresolved software dependencies. The researchers said people remain responsible for choosing research directions and evaluating evidence, particularly for open ended analysis and hypothesis generation.