codeg: Coordinate AI Coding Agents in One Workspace

Overview

Codeg is a multi-agent coding workspace that connects coding agents through the Agent Client Protocol and presents them in a shared interface. It addresses the friction of managing separate agent tools and session histories by letting developers view, search, resume, and hand off work across agents.

It is aimed at developers who already use AI coding agents and want to coordinate them around the same repositories. Codeg is available as a desktop app or as a self-hosted server, with Docker deployment and mobile clients also described in the project documentation.

Key Features

• Use a consistent conversation interface with built-in agents and compatible ACP agents.
• Search, import, and resume agent sessions, including histories created by other agents.
• Delegate work to multiple agents and view their responses alongside one another.
• Run to-dos in separate Git worktrees and review resulting changes.
• Combine conversations, files, terminals, and notes on a workspace canvas.
• Inspect and edit repository files, diffs, and Git changes beside agent conversations.
• Deploy as a desktop app, self-hosted server, or Docker container.

Use Cases

• Developers comparing one agent's implementation with another agent's review can keep both sessions in a shared workspace.
• Teams or individuals with existing agent histories can search and resume prior work without staying within one agent's interface.
• Developers handling parallel tasks can delegate work into separate worktrees and review changes before accepting them.
• Users who want to access coding sessions away from their desktop can run the server and connect supported mobile clients.

Getting Started

For desktop installers, server setup, and Docker instructions, see the README. To start the server in Docker:

docker run -d -p 3080:3080 -v codeg-data:/data ghcr.io/xintaofei/codeg:latest

Alternatives

• agent-orchestrator: Agent Orchestrator centers on Git-isolated worktrees and live tracking of agent tasks, pull requests, CI, and reviews.
• atlas: Atlas emphasizes preserving agent sessions and project context so teams can trace decisions and switch agents without restarting.
• tessera: Tessera focuses on local parallel sessions for Claude Code, Codex, and OpenCode, with Git worktrees and workflows.
• vicoa: Vicoa focuses on running agents across local and remote machines, with task and review tools available across desktop, web, mobile, and CLI.

Considerations

• Codeg connects to coding agents; users need to configure and have access to the agents and their runtime requirements.
• The project offers several deployment modes, so server and mobile use require additional setup beyond installing the desktop app.
• The repository reports 286 open issues. Review current issues and project documentation when evaluating suitability for a production workflow.