GemCP Chat integrates Gemini with extensible MCP (Model Context Protocol) servers. This repository includes Python-based MCP servers (for calculator and weather tools) that you can run alongside the Electron-based frontend. The MCP servers are designed to handle tool requests from Gemini, perform computations or data lookups, and return structured results that Gemini can render in the chat UI. To enable these servers, start the Python backend(s) and then run the Electron frontend; the app will communicate with the MCP endpoints to perform tasks like arithmetic calculations or weather queries through Gemini prompts.

Prerequisites:

• Python 3.13+ and pip
• Node.js v16+ and npm
• Optional: uv (for Python package management and running servers)

Install and run the Python MCP servers:

1. Navigate to the Python backend: cd python_backend

2. Install uv (if using uv to install dependencies): pip install uv

3. Install Python dependencies (if needed): uv pip install .

4. Set required environment variable (GOOGLE_API_KEY) in your shell:

   Linux/macOS

   export GOOGLE_API_KEY=your_api_key_here

   Windows (PowerShell)

   $env:GOOGLE_API_KEY="your_api_key_here"

5. Start each MCP server (calculator and weather): python mcp_server_calc.py python mcp_server_weather.py

6. In a separate terminal, navigate to the frontend and install dependencies: cd mcp-gemini-desktop npm install

7. Start the Electron app (development mode): npm start

8. Optional: build the Electron app for distribution: npm run build (Binaries will appear in the dist folder)

Note: You can run multiple MCP servers in parallel by invoking each Python script listed above. The Electron frontend will communicate with these servers via the configured mcp_config entries.

Tips and caveats:

• Ensure the Google API key is valid and has access to the required Gemini endpoints.
• Start the Python MCP servers before launching the Electron frontend to avoid connection errors.
• If you modify or add new MCP servers, update the mcp_config in your deployment accordingly (add new entries under mcpServers).
• The example servers include a calculator and a weather tool; you can add more by creating additional Python scripts that expose MCP-compatible endpoints.
• Check for port conflicts if you run multiple services on the same port; configure each MCP server to listen on distinct ports if needed.
• On Windows, environment variable syntax differs; use set or $env:VAR syntax, or create a .env file and load it accordingly.