Scientific papers become agentic chatbots with new tool
A new framework called Paper2Agent turns scientific papers into active AI agents. By exposing a paper’s data, code, and workflows through a Model Context Protocol server, an LLM can autonomously run demonstrations, reproduce analyses, apply methods to new data, and even collaborate with other paper agents. The aim is to boost access, reproducibility, and reuse while avoiding autonomous conclusions; agents are validated against the paper’s results to curb hallucinations. In tests across 136 papers, 74 computational-biology papers became usable agents, with failures mainly from incomplete code or missing documentation. Paper2Agent is open source on GitHub and has a live demonstration. The team plans an online platform for agent collaboration to push scientific discovery further