Paper2Agent converts research papers into callable AI agents

Paper2Agent, a framework published in Nature by Jiacheng Miao, James Zou and colleagues, converts a research paper and its codebase into an AI agent: a Model Context Protocol server that a chat agent like Claude Code can call to run the paper's tools and reproduce its analyses through natural language.

The authors' framing is that the paper stops being a document and becomes a service. "Paper2Agent addresses this challenge by converting a paper into an AI agent that functions as a virtual corresponding author, exposing its manuscript, supplementary materials, datasets, code and workflows as active, agent-native knowledge rather than static text," they write.

The case studies sit in computational biology. An agent built from AlphaGenome generated 22 tools in about 45 minutes at a cost of $14, and hit 98.7 ± 1.3% accuracy on tutorial-derived queries and 82.7 ± 2.4% on open-ended researcher questions. Additional agents based on Scanpy and TISSUE reproduced published single-cell and spatial transcriptomics results, and a multi-agent workflow prioritized a candidate causal gene for psoriasis.

At scale, the authors ran Paper2Agent across 100 computational biology papers: 74 were successfully agentified, yielding 593 validated tools and 91.2 ± 1.6% accuracy on a 300-question benchmark, with per-query cost around $0.20 and latency of 1.6 minutes.

The abstract closes on the ambition, calling the approach "a paradigm for knowledge dissemination and a collaborative ecosystem of AI co-scientists." Five researchers on our Who's Who list have shared the link.