Paper2Agent: Reimagining Research Papers As Interactive and Reliable AI Agents Authors: Jiacheng Miao, Joe R. Davis, Jonathan K. Pritchard, James Zou Submitted: 8 September 2025 Categories: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG) DOI: 10.48550/arXiv.2509.06917 --- Abstract Paper2Agent is an automated framework that transforms static research papers into interactive AI agents. This approach turns traditional research papers — often passive and complex for readers to apply — into active systems that facilitate easier reuse and scientific discovery. Key challenges Paper2Agent addresses: Traditional papers require significant effort to understand and adapt code, data, and methods. Barriers exist for dissemination and reuse of research work. The solution consists of: Automatically converting papers into knowledgeable AI assistants. Systematically analyzing the paper and codebase with multiple agents. Building a Model Context Protocol (MCP) server as the operational core. Iteratively generating and testing the MCP for reliability and robustness. Connecting MCPs with chat agents (e.g., Claude Code) enabling natural language scientific queries that invoke the original paper’s tools and workflows. Demonstrations and Validation Paper2Agent was tested on complex scientific tasks through case studies involving: AlphaGenome: Interpreting genomic variants. ScanPy and TISSUE: Single-cell and spatial transcriptomics analyses. Results show: The AI agents reproduce original paper results faithfully. They also handle novel user queries correctly, demonstrating flexibility. Significance Introduces a paradigm shift: static papers → dynamic, interactive AI agents. Paves the way for collaborative ecosystems featuring AI co-scientists. Improves knowledge dissemination and scientific productivity. --- Access and Resources PDF: View PDF version HTML (experimental): View HTML version Source: Available as TeX source files. Licensed under Creative Commons Attribution 4.0 International (CC BY 4.0) --- Associated Tools and Features on arXiv Bibliographic & Citation Tools Bibliographic Explorer Connected Papers Litmaps scite Smart Citations Code, Data, & Media Tools alphaXiv CatalyzeX Code Finder DagsHub Gotit.pub Hugging Face Papers with Code ScienceCast Demo Integrations Replicate Hugging Face Spaces TXYZ.AI Related Papers & Recommenders Influence Flower CORE Recommender About arXivLabs arXivLabs enables community collaborators to create and share experimental features, valuing openness, privacy, and community. --- Summary Paper2Agent provides an innovative framework to revolutionize scientific papers by providing AI-powered interactive agents that not only reproduce research results but also enable natural language interaction and practical downstream use. This contributes to accelerating scientific discovery, lowering the barrier to adopting new methods, and creating a rich collaborative AI co-scientist ecosystem.