MCP server for Maestro — exposes 25 workflow skills as tools, prompts, and resources for any MCP-compatible AI client.

Installation

Local (stdio)

Add to your MCP client configuration:

Claude Desktop (claude_desktop_config.json):

{
  "mcpServers": {
    "maestro": {
      "command": "npx",
      "args": ["-y", "maestro-workflow-mcp"]
    }
  }
}

Cursor (.cursor/mcp.json):

{
  "mcpServers": {
    "maestro": {
      "command": "npx",
      "args": ["-y", "maestro-workflow-mcp"]
    }
  }
}

VS Code / Antigravity (.vscode/mcp.json):

{
  "servers": {
    "maestro": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "maestro-workflow-mcp"]
    }
  }
}

Remote (HTTP)

Start as a public HTTP endpoint:

npx maestro-workflow-mcp --http --port 3001

Clients connect to http://your-server:3001/mcp. Health check at /health.

API Reference

Tools

| Tool | Description | |------|-------------| | maestro_list_commands | List all 25 commands grouped by category | | maestro_run_command | Get full instructions for a specific command | | maestro_read_context | Read .maestro/context.md or .maestro.md from a project | | maestro_init | Generate a .maestro.md template | | maestro_wave_start | Start a multi-phase wave execution | | maestro_wave_advance | Advance a wave by submitting phase output | | maestro_wave_status | Get current wave progress | | maestro_write_decision | Append a decision to .maestro/decisions.jsonl | | maestro_read_decisions | Read recent decisions from the log | | maestro_read_audit | Read audit trail with duration and cost |

Prompts

25 prompt templates — one per command. Select from your client's prompt picker:

diagnose · evaluate · reflect · refine · streamline · calibrate · fortify · zero-defect · amplify · compose · enrich · accelerate · chain · guard · iterate · temper · turbocharge · extract-pattern · adapt-workflow · onboard-agent · specialize · teach-maestro · capture · recap

Resources

| URI | Content | |-----|---------| | maestro://skill/agent-workflow | Core workflow design principles | | maestro://reference/prompt-engineering | Prompt structure, CoT, few-shot | | maestro://reference/context-management | Window optimization, state | | maestro://reference/tool-orchestration | Tool design, error handling | | maestro://reference/agent-architecture | Multi-agent topologies | | maestro://reference/feedback-loops | Evaluation, self-correction | | maestro://reference/knowledge-systems | RAG, embeddings | | maestro://reference/guardrails-safety | Validation, cost controls |

License

MIT — see LICENSE.