Keyboard-driven desktop command center that runs multiple coding agents across projects, with batch playbooks, group chat between agents, and phone control.
Running agents across projects with batch playbooks and group chat.
Core features
- Sandboxed runsNo
Maestro documents a container image for its headless Cue server, which you build and run yourself. Agents in the desktop app run on the host with no sandbox set up by Maestro.
Source - Diff review YesSource
- Agent coordination YesSource
- Scheduled tasks YesSource
- Event triggers YesSource
- Team collaboration Yes
Web Login gives each person an account on one shared Maestro with per-person attribution. There are no roles, and the desktop is the administrator.
Source - API / CLI control YesSource
Pricing
Free and open source. No pricing page or paid tier found for the app. Companion products such as Maestro Observer were not assessed.
Platforms
Desktop on macOS, Windows, Linux. Mobile is a web UI from a built-in server (QR code, optional Cloudflare tunnel), not a native app. maestro-cli for headless use.
Research notes
Codex own_subscription is the weakest value here. No Maestro doc names a ChatGPT plan. 'yes' rests on Agent Resilience docs describing Codex usage windows ('OpenAI grants those accounts reset credits that reopen a consumed window on demand'), the documented codex login flow, and the README's statement that authentication configured in Codex 'works identically in Maestro'. Downgrade to unknown if you want an explicit plan statement. Claude plan use is explicit. Integration: Codex runs through codex exec by default, with opt-in experimental app-server. Claude Code has print mode, a TUI Wrapper mode driven by maestro-p, and a Dynamic mode. The docs warn a future Claude Code change could make claude -p fail on plan logins, with the TUI Wrapper as the fallback. Docs list Hermes and Pi, which the README does not. Gemini CLI is not supported yet. docs.runmaestro.ai/remote-access returned 404; the current page is under Remote Control.