mcp-crew-ai

The MCP Crew AI Server is a Python-based implementation that runs and orchestrates CrewAI workflows using the Model Context Protocol (MCP). It automatically loads agent and task configurations from two YAML files (agents.yml and tasks.yml), enabling you to start pre-configured workflows without writing additional code. You can customize the server behavior by passing paths to these configuration files via the command line (for example, --agents path/to/agents.yml --tasks path/to/tasks.yml). The server exposes a run_workflow tool to execute pre-defined workflows and supports local development and testing in STDIO mode, which is ideal when building and debugging workflows interactively. You can also run the server via UV execution (uvx) for a streamlined development experience, for example by running uvx mcp-crew-ai --agents path/to/agents.yml --tasks path/to/tasks.yml, or start the server directly with uvx mcp-crew-ai-server to use environment-variable-based defaults. The available options include specifying the main topic (--topic), choosing a processing model (--process with sequential or hierarchical options), enabling verbose output (--verbose), and injecting additional template variables (--variables) to customize the YAML files. This allows you to easily orchestrate multi-agent interactions and generate outputs such as reports or summaries through the configured agents and tasks.

Prerequisites:

• Python 3.11+
• Pip (Python package manager)
• Git (optional, for cloning)

Option A: Install from PyPI (recommended)

1. Install the package from PyPI:

pip install mcp-crew-ai

2. Run the server (example):

mcp-crew-ai --agents path/to/agents.yml --tasks path/to/tasks.yml

Option B: Install from GitHub

1. Install directly from GitHub:

pip install git+https://github.com/adam-paterson/mcp-crew-ai.git

2. Run the server (example):

mcp-crew-ai --agents path/to/agents.yml --tasks path/to/tasks.yml

Option C: Clone and Install in editable mode

1. Clone the repository and install in editable mode:

git clone https://github.com/adam-paterson/mcp-crew-ai.git
cd mcp-crew-ai
pip install -e .

2. Run the server (example):

mcp-crew-ai --agents path/to/agents.yml --tasks path/to/tasks.yml

4. Requirements to ensure smooth operation:

• Python 3.11+
• MCP SDK
• CrewAI
• PyYAML

You can verify after installation by running the server with the required YAML configuration files as shown above.

Tips and common notes:

• The server automatically loads agents.yml and tasks.yml if you provide their paths; otherwise you can rely on environment-based defaults if supported by uvx invocation.
• You can customize server behavior with command line options like --topic, --process, and --variables to replace placeholders inside your YAML files (e.g., {topic}, {year}).
• The run_workflow tool lets you trigger pre-configured workflows defined in your tasks.yml, enabling automated multi-agent coordination.
• When using UV execution (uvx), you can execute mcp-crew-ai --agents path/to/agents.yml --tasks path/to/tasks.yml to leverage the streamlined UV runtime; uvx mcp-crew-ai-server starts the server directly in a UVX environment.
• If you encounter import or YAML parsing errors, ensure your Python environment has the required dependencies installed (MCP SDK, PyYAML, CrewAI) and that your YAML files are valid and properly formatted. Check file paths for correctness when passing --agents and --tasks.
• For production deployments, consider pinning your dependencies (requirements.txt) and using virtual environments to avoid conflicts with system Python packages.