Overview
Paper2Agent is a multi-agent system that programmatically transforms research papers and their codebases into interactive AI agents. It automates tutorial discovery, tool extraction, and MCP server generation so researchers and educators can reproduce and interact with published experiments with minimal setup.
Key features
- Automatic discovery and execution of tutorials/examples from a paper’s repository, with filtering by title or URL.
- Extraction of tutorial code into modular tools and generation of MCP servers that can be loaded by AI coding assistants like Claude Code.
- Support for local and remote MCP deployments, including Hugging Face-hosted MCP examples.
- Produces reproducible artifacts: generated MCP servers, isolated virtual environments, execution reports, and logs.
Use cases
- Reproducibility: quickly create runnable agents for research papers to reproduce results and experiments.
- Teaching & demos: package paper workflows as interactive agents for classroom use and demonstrations.
- Scientific workflow automation: extract reusable tools from large codebases and compose them into MCP services.
Technical notes
- Pipeline-driven design: tutorial scanner, tool extractor, executor, and MCP generator.
- Integrates with AI coding tools (e.g., Claude Code) for automated loading and interactive use.
- Outputs structured artifacts (
src/<repo>_mcp.py,tools/,environment directories) for engineering integration.