Reimagining research papers as interactive and reliable AI agents

Summary

Paper2Agent is an automated framework that converts a research paper into an AI agent. Rather than leaving a reader to understand and adapt the paper's code, data and methods, the system analyses the paper and its repository, builds a Model Context Protocol (MCP) server exposing the manuscript, supplementary materials, datasets, code and workflows as callable tools, and then iteratively tests the server to make it robust. The resulting agent behaves as a "virtual corresponding author": a coding agent such as Claude Code can invoke it in natural language to run the paper's workflows. The authors demonstrate it on AlphaGenome (genomic variant interpretation), TISSUE (spatial transcriptomics) and Scanpy (single-cell analysis), reporting that the agents reproduce the original papers' results and also answer new questions — including an application to psoriasis.

Contribution

Claimed: a paradigm shift for knowledge dissemination, turning the paper from a passive artefact into an active system, and a route to AI co-scientists. What the abstract supports: a working pipeline plus three case studies where the generated agents reproduced original results. The generalisation claim rests on those cases; no cross-domain benchmark is reported in the abstract.

Method

System paper with case studies. Pipeline: analyse paper and codebase → extract tools and workflows → emit an MCP server → iterative testing to harden it. Evaluation is by reproduction of the source papers' results and by novel downstream questions. The abstract does not give success rates across a corpus of papers, nor how often the extraction step fails.

Relevance to RISE

This is the first entry in the knowledge base that inverts the usual direction of the pipeline. Everything else in the catalog consumes literature (literature-discovery, literature-synthesis) or produces it (paper-drafting); Paper2Agent makes an already-published paper executable, which sits closest to replication and to the reproducibility line represented by reprorepo and social-science-replicability. It is arguably a landscape entry as much as a paper — worth considering for projects/landscape/ if the code is public and maintained. For the RISE framing it also supplies a concrete mechanism for the "Knowledge" side input: a paper-as-tool rather than a paper-as-text.

Critique / open questions

Three bioinformatics case studies is a narrow base for a claim about knowledge dissemination generally; the method depends on the paper having a working, well-structured repository, which is exactly what most papers lack — the reproducibility literature in this KB (brodeur2025reproducibility, li2026reprorepo) exists because code artefacts are usually missing or broken. Whether an agent that "reproduces the results" is checking the science or merely re-running the authors' own code is the question the abstract does not address.

Key quotes

"Conventional research papers require readers to understand and adapt the paper's code, data and methods to their work, creating barriers to...". (abstract)

"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." (abstract)