StoneX UDP Genie

You run a production-ready MCP server built on FastMCP and FastAPI to expose callable tools that AI assistants can discover and invoke. This setup focuses on a Databricks-friendly deployment path and practical local testing, so you can build, test, and connect tools quickly while keeping a clean, production-ready structure.

How to use

Start and test your MCP server locally to see how tools are discovered and called by clients. You can run the server in development mode for rapid iteration, then deploy to a Databricks App for production use. Use the provided tooling to validate health checks and user information endpoints, and to exercise each registered tool through the MCP client.

Once the server is running, you interact with it through a client that asks for available tools and then calls the specific tool endpoints. You will see a list of tools such as health checks and user information, and you can invoke each tool to verify correct behavior and data flow.

How to install

Prerequisites you need before installation include Python 3.11 or higher and a compatible runtime for running the MCP server. You will also need a Python package manager to install dependencies.

# Prerequisites check (example)
import sys
assert sys.version_info >= (3, 11)

Choose one of the installation paths below and follow the steps exactly to set up locally.

# Option 1: Using uvx (recommended) to sync and start
uv sync

# Option 2: Using pip to install dependencies in a virtual environment
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Additional sections

Configuration, security, and examples are provided to help you tailor the server for your environment. The following sections cover how to run locally, how to deploy to Databricks Apps, and how to test both locally and deployed instances.

You can customize startup behavior by choosing the appropriate command to launch the server. The project supports development scripts and a ready-to-run CLI command for local testing.

If you encounter port conflicts or import errors, ensure dependencies are installed and pick a free port for local testing. Use the port override option to avoid clashes.

Testing your MCP server involves integration tests that start the server, verify tool listing, and call each registered tool. You can also perform manual checks by invoking the health and user information endpoints through the MCP client.

Deployment to Databricks Apps enables production hosting with role-based access control and OAuth flows for user authentication. The deployment steps include syncing, deploying, and configuring external connections and app permissions, followed by testing in a Databricks-enabled environment.

Configuration snippet for MCP servers

{
  "mcpServers": [
    {
      "type": "stdio",
      "name": "stonex_udp_1",
      "command": "uv",
      "args": ["run","mcp-stonex-udp-genie"]
    },
    {
      "type": "stdio",
      "name": "stonex_udp_8080",
      "command": "uv",
      "args": ["run","mcp-stonex-udp-genie","--port","8080"]
    },
    {
      "type": "stdio",
      "name": "stonex_udp_3000",
      "command": "mcp-stonex-udp-genie",
      "args": ["--port","3000"]
    }
  ]
}

Testing and tooling

Tools available in this MCP server include health checks and a tool to fetch current user information. You can call these tools through your MCP client after listing available tools.

Available tools

health

Check server status and responsiveness to ensure the MCP server is running correctly.

get_current_user

Return information about the authenticated user making the MCP requests.