mcp -xiaozhi

This MCP server (mcp-server-xiaozhi) provides an interface to run MCP-based interactions for the Tiểu Trí project. It is driven by Python scripts and can be started to expose the MCP transport via standard IO, SSE, or HTTP as configured in mcp_config.json. The repository includes a calculator example and a generic mcp_pipe.py runner, which allows you to execute predefined tasks or communicate with the MCP pipeline to process requests.

To use, start the server with the provided Python entry (mcp_pipe.py). You can then run the calculator example as a test: python mcp_pipe.py calculator.py. You can also run all configured servers (as described in the Quick Start) to start the entire MCP pipeline. Ensure the MCP_ENDPOINT environment variable is set (or defined in an .env file) to point to your MCP instance endpoint, which the server will use to publish or fetch data as needed.

Prerequisites:

• Python 3.8+ (recommended via pyenv or system package manager)
• pip
• Access to install dependencies listed in requirements.txt

Step-by-step installation:

1. Clone or download the repository and navigate to the project directory.
2. Install Python dependencies:

pip install -r requirements.txt

3. Set up the external Phạt nguội API service if needed (refer to the Quick Start):

# Example: copy service file and enable/start the API service
sudo cp phatnguoi-api/phatnguoi-api.service /etc/systemd/system
sudo systemctl daemon-reload
sudo systemctl enable --now phatnguoi-api.service

4. Configure environment variables for MCP endpoint and any other required settings (create a .env or export variables):

export MCP_ENDPOINT=<your_mcp_endpoint>

Alternatively, place the variable in a .env file in the project root:

echo 'export MCP_ENDPOINT=<your_mcp_endpoint>' > .env

5. Run the MCP runner (as described in the Quick Start):

python mcp_pipe.py

6. If you want to test the calculator example directly:

python mcp_pipe.py calculator.py

Notes:

• Ensure the phatnguoi-api service (or any other dependent API backends) is running if your MCP tasks rely on external APIs.
• The server expects a mcp_config.json with server definitions and supports stdio, SSE, or HTTP transports depending on how you configure mcp_pipe.py and related components.
• If you change the MCP endpoint, update MCP_ENDPOINT accordingly. You can also use a .env file to persist this configuration.
• The project is a Python-based MCP server; there is no npm package here (npm_package is null).