🪨 why use many token when few token do trick — Claude Code skill that cuts 65% of tokens by talking like caveman
Desktop command center for AI agent orchestration
Maestro is a desktop app that sits in front of your coding agents and helps you run them in parallel, switch between sessions, and keep tasks organized. It works with your existing agent setup, then adds worktrees, auto-run playbooks, group chat, session history, and a command line for unattended runs.
Videos about this repo
Builders who use Claude Code, Codex, or Copilot and want a single place to coordinate many agent tasks.
You can manage parallel agent work, resume sessions, and run repeatable task batches without juggling separate terminals.
What it does
Parallel agent management
Run multiple agents and terminal sessions at once, each with its own workspace and conversation history.
Git worktrees
Create isolated worktree agents from a branch menu so parallel tasks do not collide with your main work.
Auto Run playbooks
Batch markdown checklists through agents, loop them, and track progress with full history.
Group chat
Let a moderator agent route questions between agents and synthesize answers in one conversation.
Command line interface
Use `maestro-cli` for headless runs, cron jobs, CI/CD, and JSONL output.
Session discovery
Import, search, star, rename, and resume existing sessions from supported providers.
Remote control
Monitor and control agents from a phone through the built-in web server and QR access.
Agent resilience
Retry failed turns, wait through rate limits, and optionally fail over to a backup endpoint.
How to get it
- 1Or build from source
git clone https://github.com/RunMaestro/Maestro.git cd Maestro npm install npm run dev
README
Maestro
Maestro is a cross-platform desktop app for orchestrating your fleet of AI agents and projects. It's a high-velocity solution for hackers who are juggling multiple projects in parallel. Designed for power users who live on the keyboard and rarely touch the mouse.
Collaborate with AI to create detailed specification documents, then let Auto Run execute them automatically, each task in a fresh session with clean context. Allowing for long-running unattended sessions, my current record is nearly 24 hours of continuous runtime.
Run multiple agents in parallel with a Linear/Superhuman-level responsive interface. Currently supporting Claude Code, OpenAI Codex, OpenCode, Factory Droid, and Copilot-CLI (beta) with plans for additional agentic coding tools (Gemini CLI) based on user demand.
How It Works: Maestro is a pass-through to your AI provider. Whatever MCP tools, skills, permissions, or authentication you have configured in Claude Code, Codex, or OpenCode works identically in Maestro. The only difference is we're not running interactively-each task gets a prompt and returns a response, whether it's a new session or resuming a prior one.
Features
Power Features
- 🌳 Git Worktrees - Run AI agents in parallel on isolated branches. Create worktree sub-agents from the git branch menu, each operating in their own directory. Work interactively in the main repo while sub-agents process tasks independently-then create PRs with one click. True parallel development without conflicts.
- 🤖 Auto Run & Playbooks - File-system-based task runner that batch-processes markdown checklists through AI agents. Create playbooks for repeatable workflows, run in loops, and track progress with full history. Each task gets its own AI session for clean conversation context.
- 💬 Group Chat - Coordinate multiple AI agents in a single conversation. A moderator AI orchestrates discussions, routing questions to the right agents and synthesizing their responses for cross-project questions and architecture discussions.
- 🌐 Mobile Remote Control - Built-in web server with QR code access. Monitor and control all your agents from your phone. Supports local network access and remote tunneling via Cloudflare for access from anywhere.
- 💻 Command Line Interface - Full CLI (
maestro-cli) for headless operation. List agents/groups, run playbooks from cron jobs or CI/CD pipelines, with human-readable or JSONL output for scripting. - 🚀 Multi-Agent Management - Run unlimited agents and terminal sessions in parallel. Each agent has its own workspace, conversation history, and isolated context.
- 📬 Message Queueing - Queue messages while AI is busy; they're sent automatically when the agent becomes ready. Never lose a thought.
- 🛡️ Agent Resilience - Providers fail; your turn does not have to. When a turn dies on
529 Overloadedor a spent plan quota, Maestro resends the exact prompt on its own, backing off in seconds for a blip or waiting for the real reset time it reads out of the error. One live status card replaces the wall of error dialogs, Auto Run batches resume themselves, and an agent can optionally fail over to a backup endpoint instead of waiting out the window.
Core Features
- 🔄 Dual-Mode Sessions - Each agent has both an AI Terminal and Command Terminal. Switch seamlessly between AI conversation and shell commands with
Cmd+J. - ⌨️ Keyboard-First Design - Full keyboard control with customizable shortcuts and mastery tracking that rewards you for leveling up.
Cmd+Kquick actions, rapid agent switching, and focus management designed for flow state. - 📋 Session Discovery - Automatically discovers and imports existing sessions from all supported providers, including conversations from before Maestro was installed. Browse, search, star, rename, and resume any session.
- 🔀 Git Integration - Automatic repo detection, branch display, diff viewer, commit logs, and git-aware file completion. Work with git without leaving the app.
- 📁 File Explorer - Browse project files with syntax highlighting, markdown preview, and image viewing. Reference files in prompts with
@mentions. - 🗃️ File Formats - Data files open in a viewer built for them, several with their own filtering language: jq over JSON and JSONL, sortable filtered tables for CSV and TSV, rendered Mermaid diagrams, and audio and video in a floating player.
- 🧮 Parquet Preview - Parquet files open as a live, filterable table with a schema rail. A typed query language (
ts >= now-7d and price > 100) runs against the whole file while Maestro skips the row groups it can prove cannot match, so a multi-gigabyte file filters instantly and never loads into memory. - 🔍 Powerful Output Filtering - Search and filter AI output with include/exclude modes, regex support, and per-response local filters.
- ⚡ Slash Commands - Extensible command system with autocomplete. Create custom commands with template variables for your workflows.
- 💾 Draft Auto-Save - Never lose work. Drafts are automatically saved and restored per session.
- 🔊 Speakable Notifications - Audio alerts with text-to-speech announcements when agents complete tasks.
- 🎨 Beautiful Themes - 12 themes including Dracula, Monokai, Nord, Tokyo Night, GitHub Light, and more.
- 💰 Cost Tracking - Real-time token usage and cost tracking per session and globally.
- 🏆 Achievements - Level up from Apprentice to Titan of the Baton based on cumulative Auto Run time. 11 conductor-themed ranks to unlock.
Analytics & Visualization
- 📊 Usage Dashboard - Comprehensive analytics for tracking AI usage patterns across all sessions. View aggregated statistics with multiple time ranges (day, week, month, year, all time), compare agent performance, analyze user vs. Auto Run activity distribution, and explore activity heatmaps. Fuzzy-filter your agents, then click any one to drill into its stats and see the breakdown per AI tab. Includes CSV export, real-time updates, and configurable colorblind-friendly palettes. Access via
Opt+Cmd+U(macOS) /Alt+Ctrl+U(Windows/Linux) or the Command K menu. - 🕸️ Document Graph - Visual knowledge graph of your markdown documentation. Automatically discovers internal
[[wiki-links]]and[markdown](links), visualizes document relationships with interactive nodes and edges. Toggle between force-directed and hierarchical layouts, search/filter documents, navigate via keyboard, and track external link references. Includes mini-map, legend, and pagination for large directories. Access from the File Explorer context menu or Command K menu.
Keyboard Shortcuts for Analytics Features
Usage Dashboard (Opt+Cmd+U / Alt+Ctrl+U):
| Action | Key |
|---|---|
| Navigate view tabs | Arrow Left/Right/Up/Down |
| Move between sections | Tab / Shift+Tab |
| Jump to first/last section | Home / End |
| Close dashboard | Escape |
On the Agents tab, Escape clears the agent filter first when it holds text, so filtering never costs you the dashboard.
Document Graph (Command K → "Document Graph"):
| Action | Key |
|---|---|
| Navigate to connected node | Arrow Up/Down/Left/Right |
| Cycle through connections | Tab |
| Preview document / open link | Enter |
| Recenter graph on node | Space |
| Cycle preview length | P |
| Step back / close graph | Escape |
| Search documents | Focus search input, type query |
Escape steps back one level at a time: out of the search box first (your query stays, so the matches are still highlighted while you arrow to one), then it clears the query, then it closes the graph.
Additional interactions: Drag nodes to reposition, scroll to zoom, use mini-map for overview.
Note: Maestro supports Claude Code, OpenAI Codex, OpenCode, Factory Droid, and Copilot-CLI (beta). Support for additional agents (Gemini CLI) may be added in future releases based on community demand.
Quick Start
Installation
Download the latest release for your platform from the Releases page.
Or build from source:
git clone https://github.com/RunMaestro/Maestro.git
cd Maestro
npm install
npm run dev
Requirements
- At least one supported AI coding agent installed and authenticated:
- Claude Code - Anthropic's AI coding assistant
- OpenAI Codex - OpenAI's coding agent
- OpenCode - Open-source AI coding assistant
- Copilot-CLI - GitHub's terminal coding agent (beta, multi-model via models.dev)
- Git (optional, for git-aware features)
Essential Keyboard Shortcuts
| Action | macOS | Windows/Linux |
|---|---|---|
| Quick Actions | Cmd+K | Ctrl+K |
| New Agent | Cmd+N | Ctrl+N |
| Switch AI/Terminal | Cmd+J | Ctrl+J |
| Previous/Next Agent | Cmd+[ / Cmd+] | Ctrl+[ / Ctrl+] |
| Toggle Sidebar | Cmd+B | Ctrl+B |
| New Tab | Cmd+T | Ctrl+T |
| Usage Dashboard | Opt+Cmd+U | Alt+Ctrl+U |
| All Shortcuts | Cmd+/ | Ctrl+/ |
Full keyboard shortcut reference
Screenshots
Main screen with multiple agents and conversation
Group Chat coordinates multiple AI agents in a single conversation
Quick Actions palette for rapid navigation (CTRL/CMD + K)
Git diff viewer with syntax highlighting
Documentation
Full documentation and usage guide available at docs.runmaestro.ai
- Installation
- Getting Started
- Features Overview
- Auto Run + Playbooks
- Git Worktrees
- Keyboard Shortcuts
- Context Management
- MCP Server - Connect AI apps to Maestro docs
- Troubleshooting
Community
- Discord: Join Us
- GitHub Issues: Report bugs & request features
Team
Meet the team behind Maestro at runmaestro.ai/team.
Contributing
See CONTRIBUTING.md for development setup, architecture details, and contribution guidelines.
License
Star History
Files in the repo
- .claude
- .github
- .husky
- build
- chocolatey
- docs
- e2e
- eslint-rules
- scripts
- src
- .coderabbit.yaml
- .editorconfig
- .gitignore
- .npmrc
- .prettierignore
- .prettierrc
- AGENTS.md
- ARCHITECTURE.md
- BUILDING_WINDOWS.md
- CLAUDE-AGENTS.md
- CLAUDE-CUE.md
- CLAUDE-FEATURES.md
- CLAUDE-IPC.md
- CLAUDE-PATTERNS.md
- CLAUDE-PERFORMANCE.md
- CLAUDE-PLATFORM.md
- CLAUDE-PLUGINS.md
- CLAUDE-SESSION.md
- CLAUDE-SETTINGS.md
- CLAUDE-WIZARD.md
- CLAUDE.md
- CONSTITUTION.md
- CONTRIBUTING.md
- eslint.config.mjs
- eslint.dashes.config.mjs
- LICENSE
- package-lock.json
- package.json
- PLAYBOOK_EXCHANGE_V2.md
- playwright.config.ts
- postcss.config.mjs
- PROVIDER-SUPPORT.md
- README.md
- SECURITY.md
- SYMPHONY_ISSUES.md
- SYMPHONY_REGISTRY.md
- symphony-registry.json
- tailwind.config.mjs
- THEMES.md
- tsconfig.cli.json
- tsconfig.json
- tsconfig.lint.json
- tsconfig.main.json
- vite.config.mts
- vite.config.web.mts
- vitest.config.mts
- vitest.e2e.config.ts
- vitest.integration.config.ts
- vitest.performance.config.mts
Discussion (0)
Ask about usage, or say what you built with itSign in to join the discussion.
No comments yet. Be the first to say what this is good for.
More tools
The best-benchmarked open-source AI memory system. And it's free.
Orca is the ADE for working with a fleet of parallel agents. Run any coding agent with your own subscription. Available on desktop, mobile and remote runtime.

A cross-platform desktop All-in-One assistant for Claude Code, Codex, OpenCode, OpenClaw, Grok Build & Hermes Agent. Only official website: ccswitch.io
Never stop coding. Free MIT AI gateway: one endpoint, 352 providers (150+ free), 1200+ models Kimi, Claude, GPT, Gemini, GLM, DeepSeek, MiniMax. Works with Claude Code, Codex, Cursor, OpenCode, Cline & Copilot. Quota-aware auto-fallback, RTK+Caveman compression saves 15-95% tokens, MCP/A2A, Desktop/PWA. Built by 550+ contributors
Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.