RunMaestro/Maestro

Maestro – a desktop hub for juggling many AI coding agents

What it is – Maestro is a cross‑platform desktop application that lets you run, organise, and automate dozens of AI‑powered coding assistants (Claude Code, OpenAI Codex, OpenCode, Factory Droid, Copilot‑CLI, etc.) from a single keyboard‑first interface. It treats each agent as an independent “session” with its own file system view, shell, and conversation history, while also offering tools to run them in parallel, chain them together, and keep track of cost and output.

Why it matters – Modern developers often use several AI code‑generation services at once. Switching between CLIs, browser tabs, and IDE plugins quickly becomes chaotic. Maestro centralises authentication, file access, and chat, adds robust automation (playbooks), and provides Git‑aware parallelism so you can let agents work on separate branches without stepping on each other’s toes.

Core capabilities (as described in the README)

Category	Features
Multi‑agent orchestration	Run unlimited agents side‑by‑side, each with its own AI terminal and shell terminal; a Group Chat lets a moderator AI route questions and synthesize answers across agents.
Parallel development	Git Worktrees integration creates isolated branch workspaces for each agent, enabling true parallel coding and one‑click PR creation.
Automation	Auto Run & Playbooks – a file‑system‑driven task runner that executes markdown checklists, spawning a fresh AI session per task and preserving full history.
Remote control	Built‑in web server with QR‑code access; local‑network or Cloudflare‑tunneled remote access from phones or other devices.
CLI	`maestro-cli` for headless operation, suitable for cron jobs or CI pipelines; outputs human‑readable or JSONL for scripting.
Resilience	Automatic retry on provider errors (e.g., 529 overload), optional fail‑over to backup endpoints, and queued messages while an agent is busy.
Keyboard‑first UI	Customizable shortcuts, mastery tracking, quick‑action palette (`Cmd/Ctrl+K`), and instant agent switching (`Cmd/Ctrl+J`).
File explorer & viewers	Syntax‑highlighted browsing, markdown preview, Mermaid rendering, and specialised viewers for JSON/CSV/Parquet with live filtering.
Analytics & visualisation	Usage dashboard (token/cost stats, heatmaps) and a document‑graph that maps wiki‑style links between markdown files.
Cost & token tracking	Real‑time per‑session and global token usage with cost estimates.
Gamified progress	Achievements and ranks (Apprentice → Titan) based on cumulative Auto Run time.
Theming	12 built‑in colour schemes (Dracula, Nord, Tokyo Night, etc.).

Typical workflow

Add your AI providers – authenticate Claude Code, Codex, OpenCode, etc. (Maestro simply passes through the credentials you already have).
Create a worktree – pick a Git branch; Maestro spawns a sub‑agent that works in its own directory.
Write a playbook – a markdown checklist of tasks (e.g., “generate API client → run tests → open PR”).
Run Auto Run – Maestro launches each step in a fresh AI session, queues messages, and records the full transcript.
Monitor – watch progress on the Usage Dashboard or from your phone via the remote web UI.
Iterate – use the Group Chat to ask higher‑level design questions, letting the moderator AI coordinate responses from the individual coding agents.

Getting started (from the README)

# Clone and launch the development version
git clone https://github.com/RunMaestro/Maestro.git
cd Maestro
npm install
npm run dev

Or simply download a pre‑built binary from the Releases page.

Prerequisites – at least one supported AI coding agent installed and authenticated (Claude Code, OpenAI Codex, OpenCode, Copilot‑CLI) and Git if you want the worktree features.

Who should consider using Maestro?

Power users who spend most of their time on the keyboard and need fast, mouse‑light navigation.
Developers who run multiple AI‑assisted code‑generation tools in parallel (e.g., comparing outputs from Claude Code vs. Codex).
Teams that want a shared, reproducible automation layer (playbooks) for routine AI‑driven tasks such as scaffolding, refactoring, or PR generation.
Anyone who wants a single place to monitor token usage and cost across several providers.

License & community

License: AGPL‑3.0 (requires source availability for any network‑served modifications).
Support: Discord community, GitHub Issues, and extensive docs at https://docs.runmaestro.ai.

Bottom line

Maestro is not a new AI model; it is an orchestration layer that makes existing coding assistants behave like a coordinated fleet. By combining Git worktrees, automated playbooks, a rich keyboard‑driven UI, and resilience features, it aims to turn AI‑augmented development from a series of ad‑hoc prompts into a repeatable, observable workflow.