Agi Cockpit

AGI Cockpit

**Run multiple coding agents in parallel. Keep every task, decision, and result in one cockpit.**

AGI Cockpit is a cross-platform parallel coding agent manager for coordinating Claude Code, OpenAI Codex, Cursor, Grok, and other terminal-based agents. It gives each piece of work a persistent task, lets you supervise multiple agents without mixing their context, and keeps human approvals close to the work.

This repository contains generated public documentation, screenshots, and release notes. The AGI Cockpit application source code is not included.

[Download AGI Cockpit](https://agi-labo.com/en/tools/cockpit) · [Read the official documentation](https://agi-labo.com/en/tools/cockpit/docs) · [View release notes](./RELEASES.md)


Why AGI Cockpit

• Parallel coding agent manager — Run isolated tasks side by side, see which agents are working or waiting, and return to any result without losing context.
• Multi-agent by design — Use Claude Code, OpenAI Codex, Cursor, Grok, Antigravity, or a terminal task from the same interface.
• Human decisions stay explicit — Agents can send an Ask when they need a choice or confirmation. Review and answer it from the desktop app or mobile PWA.
• Mobile PWA supervision — Check progress, answer Asks, send follow-up instructions, and operate Autoruns from another device.
• Cross-platform — Available for Windows, macOS, and Linux.
• Local working environment — Agent processes and project files remain on the computer running Cockpit. Each agent or connected service uses its own network and authentication settings.

One workspace for every agent

AGI Cockpit treats a task as the durable unit of work. A task contains the objective, working directory, selected agent, conversation, approvals, changed files, previews, and resu