companion-inc/feynman

Health v2 · maintenance only

Adoption not part of the score

9938 stars · 1097 forks observed · 2026-10-12

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

Feynman is an open-source AI research agent that reads papers, searches the web, writes cited research drafts, plans experiments, and ranks papers, usable from a terminal CLI or a local browser-based science workbench. It supports hosted and local LLM providers (LM Studio, Ollama, LiteLLM, vLLM, OpenRouter, GitHub Copilot) and can be installed as a standalone runtime bundle, npm package, or skills-only bundle for Codex.

Use cases

• run deep research with citations from papers and the web
• do a literature review on a topic
• rank papers by methodology and reproducibility
• write a research draft with cited claims
• plan and replicate ML experiments
• audit research code
• use a local LLM like Ollama for research tasks
• browse research artifacts, notebooks, and provenance in one place

When to choose

• you need an agentic research assistant with verifiable citations
• you want a terminal-first or self-hosted research workbench
• you want to use local models (Ollama, LM Studio, vLLM) for research workflows
• you need literature reviews, paper ranking, or experiment replication workflows

When to avoid

• you need a general-purpose coding agent rather than a research-focused one
• you want a fully managed cloud service with no local setup
• you need a lightweight chatbot without research tooling overhead

Facets

cli-tool · maturity active

agent-framework rag search-engine llm-inference cli chat-interface web-scraping mcp artificial-intelligence large-language-models developer-tools data-science windows cli self-hosted ai-research-agent literature-review deep-research citations science-workbench paper-ranking local-models ollama lm-studio notebooks ai-agents retrieval-augmented-generation natural-language-processing command-line macos linux nodejs web-server

Member repositories

For agents

markdown · JSON · MCP: product_card(name="Companion-Inc/feynman")