This Uber Eats MCP Server is a Python-based Proof of Concept that exposes an MCP-compliant interface for interacting with Uber Eats-like data through your LLM-powered applications. It runs via the uv (Python/uv) transport and is designed to be used with standard MCP-enabled tooling. Once running, you can query catalog-like information (restaurants, menus), retrieve item details, and perform action-oriented tasks that reflect typical MCP tool capabilities such as function invocation, intent resolution, and state maintenance across calls. The server is optimized for a lightweight stdio transport workflow, making it suitable for local development and experimentation with LLM-assisted tooling.

Prerequisites:

• Python 3.12 or higher
• An environment with internet access to install dependencies
• An Anthropic API key (or another supported LLM provider) for your agent

Setup steps:

1. Create and activate a virtual environment (example for Unix-like systems): uv venv source .venv/bin/activate # On Unix/Mac

2. Install required packages: uv pip install -r requirements.txt playwright install

3. Set your API key in a local environment file or export it: ANTHROPIC_API_KEY=your_openai_api_key_here

   or create a .env file with ANTHROPIC_API_KEY=your_openai_api_key_here

4. Run the MCP server (as shown in the README example): uv run mcp dev server.py

5. (Optional) Verify the server loads correctly by connecting an MCP client or using a debugging tool provided with your setup.

Tips and notes:

• This MCP server uses stdio as the transport, so ensure your client connects via the appropriate stdin/stdout channel.
• Set ANTHROPIC_API_KEY (or your preferred provider) before starting the server to avoid authentication errors.
• The command template for running via uvx is: uv run mcp dev server.py. If you adapt the entrypoint, update the mcp_config accordingly.
• If you encounter Playwright/browser-related issues, ensure you’ve installed the browsers via playwright install and that your environment has the necessary permissions.
• For debugging, you can extend the server with additional endpoints or log more verbose output to trace MCP calls and responses.