You run a Model Context Protocol (MCP) server that provides access to GraphQL policies, enabling you to query and orchestrate policy data efficiently from an MCP client. This guide walks you through using, installing, and configuring the server so you can integrate policy GraphQL access into your workflows with confidence.

How to use

To use the MCP server with an MCP client, first ensure your environment is prepared and your credentials are in place. Start the server locally from a directory that contains the MCP server code, then point your MCP client to the server through the local runtime. You will typically run the server in a dedicated directory and expose your GraphQL API key and endpoint via environment variables. When the server is running, your MCP client can request policies through the server’s MCP interface and receive structured policy data to inform decisions or compose responses.

How to install

# Prerequisites
- Git
- Python runtime (for the MCP server implementation)
- uv (MCP runtime) installed or available in PATH

# Clone the MCP server project
git clone https://github.com/Ad-Veritas/mcp-server-trueRAG.git
cd mcp-server-trueRAG

# Check uv availability (if not installed, install as shown below)
uv --version

Note: If uv is not installed, install it using the commands below depending on your OS.

# macOS and Linux
curl -LsSf https://astral.sh/uv/install.sh | sh

# Windows
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

Define environment variables for GraphQL access

Create a file named .env in the root of the MCP server and add:

GRAPHQL_API_KEY={your_api_key}
GRAPHQL_ENDPOINT={your_graphql_endpoint}

Start the MCP server using uv

Replace the path with the actual path to your MCP server directory

uv --directory "{path_to_mcp_server}/mcp-server-trueRAG" \
  run \
  fastmcp \
  run \
  server.py

Optional: configure an MCP client (example snippet for Claude Desktop)

"shipping-policies": {
  "command": "uv",
  "args": [
    "--directory",
    "{path_to_mcp_server}/mcp-server-trueRAG",
    "run",
    "fastmcp",
    "run",
    "server.py"
  ]
}

Additional notes

Security and access: Treat your GraphQL API key like a secret. Store it in the environment in a secure manner and avoid committing it to source control. Ensure the endpoint URL is reachable from the runtime where you start the MCP server. If you switch environments (e.g., from development to production), update the .env file accordingly and restart the MCP server to pick up the changes.

Available tools

uv

MCP runtime tool used to run the MCP server from a directory and execute the specified startup sequence.