Paper2Agent converts research papers into AI agents

A paper published in Nature on 16 September 2026 describes an automated system that turns other research papers into AI agents you can converse with. The framework is called Paper2Agent, and it targets a specific friction in academic work: readers must understand and adapt a paper's code, data and methods before they can reuse any of it.

The abstract puts its own claim plainly: "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."

Under the hood, the pipeline runs multiple agents that analyse the paper and its codebase, builds a Model Context Protocol (MCP) server, then generates and runs tests to refine that server. The resulting paper MCPs plug into a chat agent such as Claude Code, so a user invokes the paper's tools through natural language.

The authors demonstrate this on three real papers: AlphaGenome, used to interpret genomic variants, plus Scanpy and TISSUE (transcript imputation with spatial single-cell uncertainty estimation) for single-cell and spatial transcriptomics. They also chain multiple such agents together to prioritize a causal gene for psoriasis, and they validate that the agents reproduce the results of the original papers and can carry out novel user queries.

Four researchers on our Who's Who list circulated the paper.

The authors are Jiacheng Miao, Joe R. Davis, Yaohui Zhang, Jonathan K. Pritchard and James Zou.