Feynman is an open source AI research agent that runs as a command-line tool with an integrated workbench interface.
The tool addresses the need for AI-assisted research by combining natural language interaction with structured research workflows. Users type research queries or use slash commands, and the agent executes research skills—modular components that handle tasks like literature search, data analysis, and synthesis. The agent integrates with multiple language model providers including OpenAI, GitHub Copilot, OpenRouter, and local models through LM Studio, LiteLLM, Ollama, or vLLM. Installation delivers a standalone native bundle with a pinned Node.js runtime, eliminating dependency management friction.
Feynman suits researchers, engineers, and teams who want AI-assisted investigation without building custom tooling. The research skills can be installed standalone into Codex, Claude, or OpenCode environments if you need only the skill library without the full terminal application. The modular skill architecture means you can integrate Feynman's research capabilities into existing agent frameworks rather than adopting the complete terminal interface.
Development shows consistent maintenance with regular updates to the installer and model provider integrations. The project maintains clear separation between the standalone runtime bundle and the skill packages, allowing independent updates to each component. Documentation is actively maintained across installation guides, setup flows for multiple local model providers, and authentication patterns for hosted services. The codebase is written in TypeScript and includes thoughtful details like SHA-256 verification of downloaded releases and graceful handling of rate-limiting during GitHub Copilot discovery.
