Summary

Feynman is an open-source AI research agent automating literature reviews, paper analysis, and experiment replication. It targets researchers and power users, accelerating the research lifecycle with automated synthesis, verification, and replication capabilities, delivering source-grounded insights.

How It Works

Built on a Pi agent runtime and leveraging alphaXiv for paper search, Feynman uses Markdown-defined "skills" for agent behavior. Its core innovation is source-grounded outputs, linking all claims directly to original papers, documentation, or code repositories via URLs. It employs specialized agents: Researcher (evidence gathering), Reviewer (peer review simulation), Writer (drafting), and Verifier (citation validation).

Quick Start & Requirements

Installation is via a one-line script:

• macOS/Linux: curl -fsSL https://feynman.is/install | bash
• Windows (PowerShell): powershell -irm https://feynman.is/install.ps1 | iex This installs a standalone native bundle with its own Node.js runtime. Local LLM support is configured via custom providers (e.g., Ollama at http://localhost:11434/v1). Separate install-skills scripts are available, with a --repo option for project-local installs. Compute can utilize serverless GPUs (Modal) or persistent pods (RunPod). Development requires Node.js (nvm) and npm.

Highlighted Details

• Automated Workflows: Supports deep research (/deepresearch), literature reviews (/lit), peer review (/review), code audits (/audit), experiment replication (/replicate), source comparison (/compare), drafting (/draft), autonomous research (/autoresearch), and topic monitoring (/watch).
• Specialized Agents: Includes Researcher, Reviewer, Writer, and Verifier agents.
• Source Grounding: All outputs link to primary sources for verification.
• Extensibility: Integrates with AlphaXiv, Docker, web search (Gemini/Perplexity), and GPU compute providers (Modal, RunPod).

Maintenance & Community

The README provides no specific details on maintainers, community channels (Discord/Slack), or project roadmaps.

Licensing & Compatibility

Licensed under the permissive MIT License, suitable for commercial use and distribution.

Limitations & Caveats

The use of "AlphaXiv" suggests potential experimental dependencies. The README does not detail other specific limitations, unsupported platforms, or known bugs.