{
  "id": 8081540,
  "title": "Scientific papers become agentic chatbots with new tool",
  "url": "https://urgent.news/2026/09/17/scientific-papers-become-agentic-chatbots-with-new-tool",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-17T16:42:35.000Z",
  "source": {
    "name": "The Register Science",
    "slug": "the-register-science",
    "url": "https://www.theregister.com/ai-and-ml/2026/09/17/scientific-papers-become-agentic-chatbots-with-new-tool/5297276"
  },
  "original_language": "en",
  "account": "Scientific papers may soon evolve into interactive AI agents capable of discussing their findings and collaborating on new research, according to a tool developed by Stanford researchers. Paper2Agent (P2A), detailed in a Nature paper published on Wednesday, transforms scientific papers and their associated outputs into AI agents that can answer queries, reproduce analyses, apply methods to new data, and even work together on research problems. The Stanford team behind P2A envisions this as a means to accelerate the dissemination of scientific discoveries, shifting the static nature of research papers into active AI agents.\n\nJames Zou, a computer scientist and professor involved in the project, likened traditional scientific papers to \"static documents for centuries,\" emphasizing the potential of P2A to \"act as a virtual author\" with hands-on experience of a paper's work. To achieve this, P2A employs a Model Context Protocol (MCP) server that exposes a paper's tools, resources, and workflows, allowing an LLM agent to autonomously execute demonstrations, reproduce analyses, apply the paper's methods to new data, and collaborate with other paper agents. The MCP server can be hosted remotely or locally, with the latter option enabling protection of sensitive information, though such data must be excluded from the LLM backend.\n\nWhile P2A aims to enhance scientific discovery and improve access to research, it cautions against relying on the AI agents as autonomous sources of scientific conclusions, urging users to verify the accuracy of the agents' outputs. The researchers have implemented checks to prevent AI hallucinations, validating each tool used by the paper agents against the paper's results and figures. Nonetheless, users are advised to double-check the agents' work.\n\nP2A's performance has shown promise in tests involving 136 papers across three groups, with 74 successfully converted into agents. However, failures were often attributed to incomplete codebases, missing documentation, or unresolvable environment configurations. The framework is open-source and available on GitHub, with a live online version demonstrating its capabilities. Zou expressed hopes for the open-source community to contribute to improving P2A, and the team aims to develop an online platform for paper agents to collaborate and discuss their findings.",
  "summary": "Why go through the hassle of reading a study for yourself when you can turn it into an AI agent and tell it to reproduce the analysis for you?",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 2,
    "also_reported_by": [
      {
        "outlet": "The Register",
        "title": "Scientific papers become agentic chatbots with new tool",
        "url": "https://urgent.news/2026/09/17/scientific-papers-become-agentic-chatbots-with-new-tool-8084098",
        "published": "2026-09-17T16:42:35.000Z"
      }
    ]
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}