← Back to Discover

Maestro

Cross-platform desktop app for running several AI coding agents side by side, with git worktrees, markdown-driven task runs, group chat between agents and a headless CLI.

Coding agentsWorkflow automation
Visit website ↗

What it does

Maestro is an open-source desktop app for working with several AI coding agents and projects at once. It passes work through to the agents you already use, so the MCP tools, skills, permissions and authentication you set up in Claude Code, Codex or OpenCode behave the same inside it. It currently supports Claude Code, OpenAI Codex, OpenCode and Factory Droid, with Copilot-CLI, Qwen3 Coder and Oh My Pi in beta, and at least one supported agent must be installed and authenticated. Agents can run in parallel on isolated git branches through worktrees. Auto Run works through markdown checklists or a free-text goal, giving each task a fresh session, and playbooks make those runs repeatable. In Group Chat you pick one agent as moderator and it routes questions among the others over several rounds, while @mentions consult another agent inline. Other pieces include a keyboard-first interface with a terminal for each agent, session discovery that imports existing conversations, a usage dashboard with token and cost tracking, message queueing, automatic retries when a provider is overloaded or a quota runs out, remote control from a phone through a built-in web server, and a maestro-cli for headless use from cron jobs or CI pipelines. The source is under the AGPL-3.0 license. Releases can be downloaded for each platform, or the app can be built from source.

Capabilities

  • Runs multiple coding agents in parallel, each with its own workspace and history
  • Git worktrees give parallel agents isolated branches
  • Auto Run executes markdown checklists or a free-text goal, one fresh session per task
  • Group Chat with a moderator agent, plus inline @mentions to consult another agent
  • maestro-cli for headless runs from cron jobs or CI/CD pipelines
  • Usage dashboard with token and cost tracking

When to use it

  • Developers juggling several repositories or tasks with different coding agents
  • Long unattended runs driven by a written spec or checklist
  • Comparing or combining answers from more than one coding agent
  • Triggering repeatable agent playbooks from scripts or CI