Agent-Cockpit

Agent-Cockpit is an MCP server (a connector that lets AI apps use another tool) and local whiteboard that traces AI research claims back to the code that produced them.

What it is good for

  • Checking which AI research results still hold after code changes
  • Tracing a claim back to the experiment run behind it
  • Catching file changes an AI agent did not report
  • Keeping several AI agents working from one shared research board

At a glance

Who it is for
Setup difficulty
Technical ยท Guided command-line setup

Level 4 of 5. You copy and paste commands into a terminal by following the instructions. Comfortable territory for vibe coders.

Cost
Free and open source
Type
MCP server or connector
Tasks
Works with
Claude Code, Codex, Cursor, Gemini CLI, VS Code, Git

How to get started with Agent-Cockpit

  1. Open a terminal in your research folder, which must be a git repo.
  2. Run npx -y @tingyen/agent-cockpit init to write the MCP configs, AGENTS.md, and git hooks.
  3. Run npx -y @tingyen/agent-cockpit serve to start the board.
  4. Open the repo in your agent, which reads AGENTS.md and the MCP config.
  5. Try npm run demo in a clone of the repo to see a sample research project.

From the project's own documentation. If a step does not match what you see, the official pages have the latest instructions.

Before you try it

  • Requires Node.js 22.13 or newer and git.
  • The demo project also needs python3.
  • Unreported file changes block `git commit` and stop Claude Code from ending its turn.
  • The board runs locally on 127.0.0.1 and its checks make no LLM calls.

Momentum

Rank this week
Not ranked (last #78, week 45, 2026)
Best rank
#78 week 45, 2026
GitHub stars
501

GitHub numbers as of Nov 16, 2026. Updated weekly.

75 304 533 Oct 31, 2026 Nov 16, 2026
GitHub stars for Agent-Cockpit: 107 on Oct 31, 2026, 501 on Nov 16, 2026. Source: GitHub.

Ranking history

First ranked in week 45, 2026.

Questions

Is Agent-Cockpit free?
Yes. It is free and open source (AGPL-3.0 license).
Do I need to know how to code to use Agent-Cockpit?
You do not need to write code, but setup means copying and pasting commands into a terminal by following its instructions.
What does Agent-Cockpit work with?
According to its own documentation, it works with Claude Code, Codex, Cursor, Gemini CLI, VS Code, and Git.
How do I get started with Agent-Cockpit?
1. Open a terminal in your research folder, which must be a git repo. 2. Run npx -y @tingyen/agent-cockpit init to write the MCP configs, AGENTS.md, and git hooks. 3. Run npx -y @tingyen/agent-cockpit serve to start the board. 4. Open the repo in your agent, which reads AGENTS.md and the MCP config. 5. Try npm run demo in a clone of the repo to see a sample research project.
What should I know before trying Agent-Cockpit?
Requires Node.js 22.13 or newer and git. The demo project also needs python3. Unreported file changes block `git commit` and stop Claude Code from ending its turn. The board runs locally on 127.0.0.1 and its checks make no LLM calls.
What is Agent-Cockpit good for?
Checking which AI research results still hold after code changes, tracing a claim back to the experiment run behind it, catching file changes an AI agent did not report, and keeping several AI agents working from one shared research board.

Similar agents to Agent-Cockpit

  • codebase-memory-mcp

    Maps a code repository so agents can find code faster, rather than tracking research claims and experiment runs.

  • GitDiagram

    Turns a GitHub repository into an architecture diagram instead of tracking evidence for research results.

  • SoloMD

    A markdown note editor that agents can read and edit, with no checking of claims against code.

  • Argo

    Searches many sources for agents doing web research, rather than verifying results the agent produced itself.

  • GitHub MCP Server

    Lets AI tools manage GitHub repositories, issues, and pull requests instead of tracking research evidence.

  • OmO

    OmO is a coding agent app that turns a single prompt into finished software work for developers and builders who want agents to handle multi-step jobs.

Last updated Nov 16, 2026. Tracked since Oct 31, 2026. Spotted an error? Tell us.