v1.2.0 — MCP Server

🚀 MCP Server — maestro-workflow-mcp

Maestro is now available as a live Model Context Protocol server. Any MCP-compatible client can connect — no file copying required.

Install

npm: maestro-workflow-mcp

Local (stdio) — add to your MCP client config:

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

Remote (HTTP) — host as a public endpoint:

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

---

What's New

MCP Server (mcp-server/)

• 20 Prompt Templates — one per command, selectable from the prompt picker
• 4 Tools:
  • maestro_list_commands — browse all commands by category
  • maestro_run_command — get full instructions for any command
  • maestro_read_context — read .maestro.md from a project path
  • maestro_init — generate a .maestro.md context template
• 8 Resources — core agent-workflow skill + 7 domain references (prompt engineering, context management, tool orchestration, agent architecture, feedback loops, knowledge systems, guardrails & safety)
• Dual Transport — stdio for local clients, Streamable HTTP for public/remote access
• Health Check — GET /health endpoint for monitoring in HTTP mode
• Build-time Bundling — all 21 skills + 7 references compiled into a single TypeScript module via scripts/bundle-skills.js

Infrastructure

• New scripts/bundle-skills.js — parses YAML frontmatter from all source skills and generates typed skill data
• Updated .gitignore — excludes MCP server build artifacts
• Updated README.md — MCP Server section with installation instructions and project structure tree

Compatibility

Works with any MCP-compatible client:

• Claude Desktop
• Cursor
• VS Code / Antigravity
• Gemini CLI
• Any Streamable HTTP MCP client

Full Changelog: v1.1.0...v1.2.0