freshcrate
Skin:/
Home > AI Agents > maestro-orchestrate

maestro-orchestrate

Multi-agent orchestration platform for Gemini CLI and Claude Code — 22 specialists, parallel subagents, persistent sessions, and built-in code review, debugging, security, SEO, accessibility, and comp

Why this rank:Strong adoptionRelease freshnessHealthy release cadence

Description

Multi-agent orchestration platform for Gemini CLI and Claude Code — 22 specialists, parallel subagents, persistent sessions, and built-in code review, debugging, security, SEO, accessibility, and compliance tools

README

Maestro

Version License Gemini CLI Claude Code Codex Qwen Code

Maestro is a multi-agent development orchestration platform with 39 specialists, an Express path for simple work, a 4-phase standard workflow for medium and complex work, persistent session state, and standalone review/debug/security/perf/seo/accessibility/compliance entrypoints. It runs from one canonical src/ tree across Gemini CLI, Claude Code, Codex, and Qwen Code.

Runtime Targets

Runtime Location Public Surface Notes
Gemini CLI repo root /maestro:* Snake-case agents, TOML commands, hooks, TOML shell policies
Claude Code claude/ /orchestrate, /review, ... Kebab-case agents with maestro: subagent names
Codex plugins/maestro/ $maestro:* Plugin skills, spawn_agent, no runtime hooks
Qwen Code qwen/ /maestro:* Gemini-CLI-compatible extension, QWEN.md context file, SubagentStart/SubagentStop hooks

Getting Started

Prerequisites

  • One supported runtime: Gemini CLI, Claude Code, Codex, or Qwen Code
  • Node.js 18+ for the MCP server and helper scripts
  • Gemini CLI and Qwen Code only: enable experimental subagents in ~/.gemini/settings.json (Gemini) or ~/.qwen/settings.json (Qwen)
{
  "experimental": {
    "enableAgents": true
  }
}

Maestro does not edit ~/.gemini/settings.json or ~/.qwen/settings.json for you.

Installation

Gemini CLI

gemini extensions install https://github.com/josstei/maestro-orchestrate

Local development:

git clone https://github.com/josstei/maestro-orchestrate
cd maestro-orchestrate
gemini extensions link .

Verify with gemini extensions list.

Claude Code

Marketplace install:

claude plugin marketplace add josstei/maestro-orchestrate
claude plugin install maestro@maestro-orchestrator --scope user

Development / temporary loading:

git clone https://github.com/josstei/maestro-orchestrate
claude --plugin-dir /path/to/maestro-orchestrate/claude

More Claude-specific setup and plugin management lives in claude/README.md.

Codex

codex marketplace add josstei/maestro-orchestrate

Then start Codex, run /plugins, and select Maestro → Install.

Local development:

git clone https://github.com/josstei/maestro-orchestrate
codex marketplace add /path/to/maestro-orchestrate

More Codex-specific setup and runtime details live in plugins/maestro/README.md and docs/runtime-codex.md.

Qwen Code

qwen extensions install https://github.com/josstei/maestro-orchestrate

Local development:

git clone https://github.com/josstei/maestro-orchestrate
cd maestro-orchestrate
qwen extensions link .

Verify with qwen extensions list. Qwen Code uses the same /maestro:* command surface as Gemini CLI and reads QWEN.md as its context file.

Quick Start

Start a full orchestration with the runtime-specific entrypoint:

Runtime Example
Gemini CLI /maestro:orchestrate Build a REST API for a task management system with user authentication
Claude Code /orchestrate Build a REST API for a task management system with user authentication
Codex $maestro:orchestrate Build a REST API for a task management system with user authentication
Qwen Code /maestro:orchestrate Build a REST API for a task management system with user authentication

Maestro classifies the task, chooses Express or Standard workflow, asks the required design questions, produces an implementation plan when needed, delegates execution to specialists, runs a quality gate, and archives the session state in docs/maestro/.

Commands

Capability Gemini CLI Claude Code Codex
Orchestrate /maestro:orchestrate /orchestrate $maestro:orchestrate
Execute /maestro:execute /execute $maestro:execute
Resume /maestro:resume /resume $maestro:resume-session
Status /maestro:status /status $maestro:status
Archive /maestro:archive /archive $maestro:archive
Review /maestro:review /review $maestro:review-code
Debug /maestro:debug /debug $maestro:debug-workflow
Security Audit /maestro:security-audit /security-audit $maestro:security-audit
Performance Check /maestro:perf-check /perf-check $maestro:perf-check
SEO Audit /maestro:seo-audit /seo-audit $maestro:seo-audit
Accessibility Audit /maestro:a11y-audit /a11y-audit $maestro:a11y-audit
Compliance Check /maestro:compliance-check /compliance-check $maestro:compliance-check

For Codex, Maestro intentionally avoids bare skill names that collide with host commands. Use $maestro:review-code, $maestro:debug-workflow, and $maestro:resume-session so Codex's built-in /review, /debug, and /resume commands keep working.

Qwen Code uses the same /maestro:* command surface as Gemini CLI.

Workflow

  • Express: For simple work. Maestro asks 1-2 clarifying questions, proposes a brief, delegates to one specialist, runs code review, and archives without a design doc or implementation plan.
  • Standard: For medium and complex work. Maestro runs Design, Plan, Execute, and Complete phases with explicit approval gates, phased execution, and final review blocking on unresolved Critical or Major findings.

Documentation

License

Apache-2.0

Release History

VersionChangesUrgencyDate
v1.6.4 ### Added - **Maestro cheatsheet**: added an English quick-reference guide covering supported runtimes, common commands, workflow concepts, and suggested reading order. ### Changed - **npm package identity**: renamed the planned npm package to `@josstei/maestro`, added `hello@josstei.dev` to public author metadata, and moved the stable release publish path into GitHub Actions with npm token authentication. ### Fixed - **Stable npm release recovery**: Release now uses `NPM_TOKEN` for stableHigh4/30/2026
v1.6.3 ### Added - **5 new MCP tools** (Session pack 5 → 10; total pack count 12 → 17): - `enter_design_gate`, `record_design_approval`, `get_design_gate_status` — server-side design-gate lifecycle; `create_session` now enforces approval before session creation - `scan_phase_changes`, `reconcile_phase` — post-hoc phase reconciliation when an agent's actual file manifest diverges from the plan - **17 additional specialist agents** — Expanded the roster from 22 to 39 specialists: - **ML / AI**: `High4/21/2026
v1.6.2 ### Added - **Qwen Code runtime support** (#26) — Fourth runtime target with hook adapter, extension manifest (`qwen-extension.json`), context file (`QWEN.md`), 22 agent stubs, and `hooks.json` for `SubagentStart`/`SubagentStop` lifecycle events - **`getExitCode(result)` adapter contract** (#42) — Each hook adapter now owns its exit code semantics; shared `exit-codes.js` constants (`EXIT_SUCCESS`, `EXIT_BLOCK`) verified against Gemini CLI, Claude Code, and Codex CLI source - **npm publishing wHigh4/15/2026
v1.6.1 ### Added - **Self-contained Claude runtime payload** (`src/manifest.js`) — Added `'claude'` to the self-contained runtime payload manifest rule (already used by Codex) so the full `src/` tree is bundled into `claude/src/`; canonical-source resolution now succeeds from plugin caches outside the repo tree - **`withIsolatedClaudePlugin` test helper** (`tests/integration/helpers.js`) — Copies the generated `claude/` directory into an isolated temp directory for testing plugin behavior outside theHigh4/10/2026
v1.6.0 ### Added - **Codex runtime target** (`plugins/maestro/`) — Complete third runtime for OpenAI Codex with plugin manifest, 22 agent stubs, 19 skills, MCP entry-point, and runtime guide; agents use kebab-case names and `spawn_agent(...)` delegation - **Canonical source architecture** — Single `src/` tree serves all three runtimes (Gemini, Claude, Codex) via `requireFromCanonicalSrc()` dynamic resolution, eliminating verbatim lib copies and runtime-specific source directories - **`get_agent` MCP Medium4/10/2026
v1.5.0 ### Added - **`get_skill_content` MCP tool** — Reads delegation protocols, templates, and reference documents by identifier via MCP, bypassing workspace sandbox restrictions. Used by the orchestrate command to load non-skill resources (methodology skills are loaded via `activate_skill`). - **`references/orchestration-steps.md`** — Shared numbered-step sequence (40 steps with inline HARD-GATEs) loaded by both Gemini CLI and Claude Code orchestrate commands as the sole procedural authority. - **High4/4/2026
v1.4.0 ### Added - **10 new specialist agents** — `seo_specialist`, `copywriter`, `content_strategist`, `ux_designer`, `accessibility_specialist`, `product_manager`, `analytics_engineer`, `i18n_specialist`, `design_system_engineer`, `compliance_reviewer`; roster expanded from 12 to 22 - **MCP server** (`mcp/maestro-server.js`) — Bundled Model Context Protocol server registered via `mcpServers` in `gemini-extension.json` with 9 tools: `initialize_workspace`, `assess_task_complexity`, `validate_plan`, Low3/19/2026
v1.3.0 ### Added - **Plan-based execution mode recommendation** — When `MAESTRO_EXECUTION_MODE=ask` (default), the orchestrator analyzes the implementation plan's dependency graph and presents a data-driven parallel vs sequential recommendation via `ask_user` - **Execution mode gate enforcement** — `<HARD-GATE>` language in the execution skill ensures the mode prompt cannot be skipped; safety fallback stops delegation if `execution_mode` is missing from session state - **Mandatory gate references acrLow3/7/2026
v1.2.1 ### Added - **Expanded test coverage** — Added 91 unit tests and migrated integration tests to Node.js to validate hooks, dispatch, state handling, config resolution, and timeout behavior - **Cross-platform PR CI matrix** — Added GitHub Actions workflow (`.github/workflows/ci.yml`) running `node tests/run-all.js` on both `ubuntu-latest` and `windows-latest` ### Changed - **Cross-platform runtime migration** — Replaced bash/Python hook and script execution paths with Node.js entry points for Low2/23/2026
v1.2.0### Added - **Hooks-based lifecycle middleware** — BeforeAgent and AfterAgent hooks with shared shell library (`hooks/lib/common.sh`), `safe_main` wrapper for guaranteed JSON output, and advisory error handling - **Agent tracking** — BeforeAgent/AfterAgent hooks track active agent identity via `/tmp/maestro-hooks/<session-id>/active-agent`; lazy state creation on first write, stale-pruned during BeforeAgent - **Handoff report validation** — AfterAgent hook validates delegated agent output iLow2/19/2026
v1.1.1 ### Fixed - Removed extension settings prompts from install — Gemini CLI doesn't support default values, so users were forced through 13 prompts on install. All settings now use orchestrator defaults and are configurable via environment variables. ### Changed - README configuration section renamed from "Extension Settings" to "Environment Variables" with all 13 parameters documentedLow2/16/2026
v1.1.0 ### Added - Extension settings with 10 configurable parameters via environment variables - Maestro branded dark theme with warm gold accents - Shell-based parallel dispatch for concurrent subagent execution (`scripts/parallel-dispatch.sh`) - Agent base protocol with pre-flight procedures and structured output formatting - Settings references in delegation, execution, session-management, and validation skills - TechLead orchestrator startup checks with settings resolution - Filesystem safety prLow2/16/2026

Dependencies & License Audit

Loading dependencies...

Similar Packages

pickle-rick-claudeđŸĨ’ Pickle Rick for Claude Code — autonomous PRD-driven coding loops + relentless code review. Ralph Loop toolkit.v1.88.0
be-my-butlerOrchestrate multiple agents to execute Claude Code workflows with cross-model verification for reliable AI code automation.main@2026-06-05
hermes-agentThe agent that grows with youv2026.6.5
.claudeThe Ultimate Claude Code Toolkit — 127 skills, 86 agents, 109 marketplace repos (11,700+ community skills), 30 commands, 8 hooks, GSD framework. Drop-in ~/.claude config that auto-activates the right master@2026-06-03
runtmOpen-source sandboxes where coding agents build and deploy. Spin up isolated environments where Claude Code, Cursor, and other agents code and deploy software.main@2026-05-25

More in AI Agents

hermes-agentThe agent that grows with you
awesome-copilotCommunity-contributed instructions, agents, skills, and configurations to help you make the most of GitHub Copilot.
CopilotKitThe Frontend Stack for Agents & Generative UI. React + Angular. Makers of the AG-UI Protocol
e2bE2B SDK that give agents cloud environments