Maestro

// Overview

Maestro is a cross-platform desktop app for orchestrating a fleet of AI coding agents and projects at once. It supports Claude Code, OpenAI Codex, OpenCode, Factory Droid and Copilot CLI (beta), and it is a pass-through to whichever provider you already use: the MCP tools, skills, permissions and authentication configured in those agents work identically inside Maestro. The difference is that Maestro drives them non-interactively — each task gets a prompt and returns a response, in a new session or resuming a prior one.

The workflow it is built around is spec-then-execute. You collaborate with an agent to write a detailed specification document, then Auto Run works through it automatically, giving each task a fresh session with clean context so long unattended runs do not drift. Playbooks make that repeatable: file-system-based markdown checklists batch-processed through agents, runnable in loops with full history. Git worktrees keep parallel work from colliding — sub-agents are created from the git branch menu and each operates in its own directory, so you can work interactively in the main repo while they run, then open PRs from the UI.

The rest is the scaffolding that makes a fleet manageable. Every agent has both an AI terminal and a command terminal, switched with one shortcut. Group Chat puts several agents in one conversation with a moderator agent routing questions and synthesising answers. A built-in web server with QR-code access (and optional Cloudflare tunnelling) turns a phone into a remote control, while maestro-cli covers headless operation from cron or CI. Agent resilience retries a turn that dies on a 529 or a spent quota, backing off for a blip or waiting out the real reset window, so Auto Run batches resume themselves. Maestro is AGPL-3.0 licensed.

// What it does

• Runs Claude Code, OpenAI Codex, OpenCode, Factory Droid and Copilot CLI (beta) side by side, as a pass-through to your own provider config
• Git worktree sub-agents — parallel agents on isolated branches in their own directories, with one-click PR creation
• Auto Run and Playbooks: markdown task checklists batch-processed through agents, each task in a fresh session, runnable in loops
• Group Chat, where a moderator agent routes questions across several agents and synthesises their answers
• Mobile remote control through a built-in web server with QR access, plus Cloudflare tunnelling for access from outside the LAN
• maestro-cli for headless use — list agents and groups, run playbooks from cron or CI, with human-readable or JSONL output
• Agent resilience: automatic retry and backoff on 529 Overloaded or exhausted plan quota, with optional failover to a backup endpoint
• Session discovery that imports existing conversations from every supported provider, plus git diffs, a file explorer, Parquet preview, a usage dashboard and cost tracking

// Getting started

Maestro is a desktop app. Grab a packaged release, or build from source if you want to track the repo.

git clone https://github.com/RunMaestro/Maestro.git
cd Maestro
npm install
npm run dev

Cmd+K / Ctrl+K   Quick Actions
Cmd+N / Ctrl+N   New agent
Cmd+J / Ctrl+J   Switch AI / terminal
Cmd+/ / Ctrl+/   All shortcuts

Commands and code are distilled from the project's own documentation — always check the official repo for the latest.

// When to use it

• Keep several repos or features moving at once instead of babysitting one agent session at a time
• Turn a written spec into an unattended Auto Run batch, with each task starting from clean context
• Run agents on isolated git worktrees so parallel work never fights over the same working tree
• Kick off or check on a long agent run from your phone, or schedule a playbook from cron

// How Maestro compares

Maestro alongside other open-source coding-agent harnesses & runtimes tools AI/TLDR tracks, ranked by GitHub stars.