Introduction

Feynman is an open source AI research agent CLI for searching papers, synthesizing evidence, verifying citations, and replicating experiments.

Feynman is an open source AI research agent that runs in your terminal. It searches academic papers, synthesizes evidence across sources, verifies citations, and replicates experiments — all from a conversational interface built on Pi and alphaXiv.

Every output is source-grounded. Claims link to papers, docs, or repositories with direct URLs. You can ask naturally in free-form language or use slash commands to invoke specific research workflows.

Research subagents

Feynman ships four bundled research subagents that are dispatched automatically depending on the workflow:

• Researcher — gathers evidence across papers, the web, repositories, and documentation
• Writer — produces structured paper-style drafts from research notes
• Reviewer — runs simulated peer review with severity-graded feedback and a revision plan
• Verifier — checks inline citations, verifies source URLs, and cleans up dead links

The lead agent plans and delegates work; subagents handle decomposed subtasks and hand off results via files.

Research workflows

Ask Feynman in natural language or use slash commands as shortcuts. The table below lists the built-in workflows:

Command	Description
/deepresearch <topic>	Source-heavy multi-agent investigation with parallel researchers, synthesis, and verification
/lit <topic>	Literature review with consensus findings, disagreements, and open questions
/review <artifact>	Simulated peer review with severity ratings and a revision plan
/audit <item>	Compares paper claims against the public codebase to surface mismatches
/replicate <paper>	Replicates experiments on local or cloud GPUs
/compare <topic>	Side-by-side source comparison matrix
/draft <topic>	Paper-style draft from research findings
/autoresearch <idea>	Autonomous experiment loop
/watch <topic>	Recurring research watch on an evolving topic
/outputs	Browse all research artifacts (papers, outputs, notes)

How it works

Feynman is built on Pi for the agent runtime and alphaXiv for paper search, Q&A, code reading, and annotations. Research capabilities are delivered as Pi skills — Markdown instruction files synced to ~/.feynman/agent/skills/ on startup.

Additional tools include Docker for isolated container execution, Gemini or Perplexity for zero-config web search, Modal for serverless GPU compute, and RunPod for persistent GPU pods.

Get started

Installation

Install Feynman on macOS, Linux, or Windows using the one-line installer, pnpm, or bun.

Setup

Configure your model provider, API keys, and alphaXiv authentication.

Quick Start

Launch the REPL, run your first research query, and explore slash commands.

CLI Commands

Full reference for all Feynman CLI commands and flags.