🚀 CrewAI MCP Orchestrator

🚀 CrewAI MCP Orchestrator

The CrewAI MCP Orchestrator is a highly capable Model Context Protocol (MCP) server that transforms any compatible LLM or AI Assistant into a master orchestrator of the CrewAI framework.

This server provides 13 dynamic tools, integrated RAG documentation search (ChromaDB), and programmatic control over CrewAI projects, enabling LLMs to dynamically generate, edit, test, and execute multi-agent systems.

🌟 Key Features

1. Integrated Knowledge Base (RAG): Automatically chunks and indexes CrewAI Markdown documentation so the AI can read guides and concepts via crewai_query_knowledge.
2. Project Scaffolding: Create native CrewAI projects and flows isolated in a /workspace directory.
3. YAML Lifecycle Management: Define agents, tasks, and flows directly using Python and Pydantic without breaking the CLI structures.
4. Observability & Debugging: Automatically run tests, train models, and replay failed tasks.
5. Multi-Transport Support: Works locally via stdio for IDEs or via SSE/HTTP using the included Docker configuration.

🛠️ Installation & Setup

We recommend using uv for lightning-fast dependency management.

# 1. Clone the repository
git clone https://github.com/tu-usuario/cwai-mcp.git
cd cwai-mcp

# 2. Sync and install dependencies (creates isolated .venv)
uv sync

# 3. Add documentation (Optional but recommended)
# Place CrewAI markdown files inside the /docs folder to be automatically indexed.

🔌 IDE & Client Integration (MCP)

To connect your favorite AI IDE or Assistant to this server, you need to configure an MCP connection over stdio. Since the server uses uv and its own virtual environment, it's highly recommended to point directly to the .venv Python executable for speed and clean stdout streams.

🌌 Antigravity

Open your mcp_config.json (usually located in .gemini/config/mcp_config.json) and add:

{
  "mcpServers": {
    "crewai-orchestrator": {
      "command": "C:\\Ruta\\Absoluta\\cwai-mcp\\.venv\\Scripts\\python.exe",
      "args": ["-X", "utf8", "-m", "crewai_mcp.server"],
      "cwd": "C:/Ruta/Absoluta/cwai-mcp",
      "env": {
        "CREWAI_MCP_TRANSPORT": "stdio",
        "PYTHONUTF8": "1"
      }
    }
  }
}

🤖 Claude Desktop / Claude Code

Open your Claude Desktop config file (Windows: %APPDATA%\Claude\claude_desktop_config.json, Mac: ~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "crewai-orchestrator": {
      "command": "/absolute/path/to/cwai-mcp/.venv/bin/python",
      "args": ["-m", "crewai_mcp.server"],
      "cwd": "/absolute/path/to/cwai-mcp",
      "env": {
        "CREWAI_MCP_TRANSPORT": "stdio"
      }
    }
  }
}

💻 Roo Code / Cline (VS Code)

Open your MCP settings from the extension UI or edit cline_mcp_settings.json:

{
  "mcpServers": {
    "crewai-orchestrator": {
      "command": "C:\\Ruta\\Absoluta\\cwai-mcp\\.venv\\Scripts\\python.exe",
      "args": ["-X", "utf8", "-m", "crewai_mcp.server"],
      "cwd": "C:/Ruta/Absoluta/cwai-mcp",
      "env": {
        "CREWAI_MCP_TRANSPORT": "stdio",
        "PYTHONUTF8": "1"
      }
    }
  }
}

🐾 OpenClaw / Generic Stdio

For OpenClaw or any other generic MCP client, simply configure a local process executing:

• Command: /absolute/path/to/.venv/bin/python (or python.exe on Windows)
• Args: ["-m", "crewai_mcp.server"]
• Env: CREWAI_MCP_TRANSPORT=stdio

🐳 Docker Deployment (SSE)

If you wish to deploy the MCP orchestrator as a standalone microservice (e.g., inside a Docker Triad architecture), it supports Server-Sent Events (SSE).

# Build and run using Docker Compose
docker-compose up --build -d

The server will be available at http://localhost:8000/sse.

🧰 Available Tools

📝 License

MIT License. Created to supercharge AI-driven multi-agent orchestration.