Maestro: an orchestrator of 39 specialists
Maestro
An orchestration plugin for Claude Code, Codex, Gemini CLI and Qwen Code: 39 specialists, parallel subagents, persistent sessions and standalone audits
- 485 ★ on the repository
- Last commit
- Apache-2.0
Broad access
Why this level
- Specialists run in parallel and change code in the project, and execution follows the plan without you in every step
- The MCP server and scripts run locally, and sessions accumulate files in docs/maestro
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/maestroDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Specialists run in parallel and change code in the project, and execution follows the plan without you in every step
- The MCP server and scripts run locally, and sessions accumulate files in docs/maestro
This is third-party code. Review the repository files before installing.
What it does
Maestro splits a task among specialists. Simple work has a fast Express path: one or two clarifying questions, a brief, one executor, code review and an archive. Medium and complex tasks follow a standard four-phase process of design, plan, execution and completion, with explicit approval gates. The specialists sit in the agents directory: architect, coder, tester, debugger, security engineer, SEO specialist, accessibility specialist, technical writer and others, including rare ones such as a COBOL engineer and z/OS. Session state, design documents, plans and phase reports are stored in docs/maestro and let you resume work. Review, debugging, a security audit, a performance check, an SEO audit, an accessibility check and a compliance check also run on their own. One src/ codebase builds for four environments: Gemini CLI, Claude Code, Codex and Qwen Code, each with its own set of commands.
Who it is for. For developers who give agents large tasks and want a plan with approval gates, parallel executors and resumable sessions.
Good fit when
- A task spans several areas, such as API, database and UI, and is worth splitting among specialists
- You need a plan and a design document with explicit approval before the agent changes code
- You need a standalone security, accessibility, performance or SEO audit of existing code
- The work is long and session state should persist on disk so a resume command can pick it up
- You switch between Claude Code, Codex, Gemini CLI and Qwen Code and want the same set of specialists
Not a fit when
- The task takes five minutes: even the Express path asks clarifying questions and keeps an archive
- You cannot run many subagents: cost and token use grow with the number of executors
- The repository cannot tolerate files in docs/maestro: the state directory is changed with MAESTRO_STATE_DIR, but a folder is always needed
- You need one skill, such as review only: a separate tool without the full process fits better
Example request
Run /orchestrate: build a REST API for a task list with user registration and show the design and plan for approval firstLimitations
- One of the four supported agents and Node.js 20 or newer for the MCP server and helper scripts are required.
- Gemini CLI and Qwen Code need experimental subagents enabled in settings.json, and the plugin does not edit that file itself.
- Claude Code and Codex use different command names to avoid clashing with built-ins: use /review-code, /debug-workflow, /resume-session. Codex has no runtime hooks.
- Maestro follows the permissions and sandbox of the chosen agent, and its specialists can be limited with MAESTRO_DISABLED_AGENTS.
- The license is Apache-2.0.
How to disable. In Claude Code run claude plugin uninstall maestro or disable the plugin in the /plugin menu. Gemini CLI and Qwen Code have an extensions uninstall command, and Codex has the /plugins menu. The session archive lives in docs/maestro and is deleted by hand.
FAQ
How do I install Maestro in Claude Code?
Run claude plugin marketplace add josstei/maestro-orchestrate, then claude plugin install maestro@maestro-orchestrator --scope user. The main command is /orchestrate, plus /review-code, /debug-workflow, /security-audit, /perf-check, /seo-audit, /a11y-audit and /compliance-check.
How does the Express path differ from the standard one?
Express is for simple work: clarification, a brief, one specialist, review and an archive with no design document. The standard path goes through four phases with approval gates and closes with a review that leaves no unresolved critical findings.
Where does Maestro keep session state?
In docs/maestro, which MAESTRO_STATE_DIR changes. It holds the active session, design, plan, phase reports and the archive.
Which settings does Maestro have?
MAESTRO_EXECUTION_MODE (ask, parallel or sequential), MAESTRO_VALIDATION_STRICTNESS (strict, normal, lenient), MAESTRO_AUTO_ARCHIVE, MAESTRO_MAX_RETRIES defaulting to 2, MAESTRO_MAX_CONCURRENT and MAESTRO_DISABLED_AGENTS.
What must be enabled in Gemini CLI before installing?
Experimental subagents: set experimental.enableAgents to true in ~/.gemini/settings.json. For Qwen Code the same setting lives in ~/.qwen/settings.json.
Discussion
No questions or reviews yet.
Related
PluginsA skill with 25 commands to audit and improve the prompts, tools and agent designs of a project in Cursor, Claude Code and Codex
Multi-agent orchestration for Claude Code: teams, roles, autopilot and task routing across skills and subagents
A desktop app where Claude Code, Codex, Gemini and other agents work as a team of specialists in a shared project
Workflows for Claude Code: /do, /omo, /bmad-pilot and more commands, with Codex, Gemini or OpenCode writing the code through codeagent-wrapper
A layer over coding agents that assembles a temporary specialist team per task inside Claude Code, Codex, Cursor and Gemini
Orchestration layer for Claude Code, Cursor and Codex: a board of 495 expert personas gives a verdict with named dissent, plus hundreds of skills
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Repository stars | Main difference |
|---|---|---|---|
| Maestro: an orchestrator of 39 specialists (this entry) | no | 485 | |
| CLI Agent Orchestrator | no | 1,489 | an AWS CLI where a supervisor delegates to other agents, without a catalog of 39 specialists |
| Zeroshot | no | 2,063 | an agent graph: one writes, independent ones verify and failures go back |
| Ruflo | no | 80,113 | agent swarms and background workers with memory; Maestro uses file-based sessions |
README in short
The README presents Maestro as a multi-agent orchestration platform with 39 specialists and a table of target environments. Then come requirements, installation for each environment, a quick start with a sample request and a settings table. After that it collects commands for the four environments, describes the Express and standard processes, output files with success criteria, a security section and links to docs: overview, architecture, examples.