Happy slice of mcp pizzaMCP.pizza
gemini-desktop

gemini-desktop

MCP.Pizza Chef: kkrishnan90

A desktop app for Mac and Windows where you chat with Gemini and attach extra tools, with each tool call shown as it succeeds or fails. In practice it cannot hold a conversation today: the four models it offers are fixed in the source and every one has been shut down by Google, including Gemini 2.0 Flash in June 2026, and the app rejects any name outside that list. The setup notes are wrong too — the wrong key name, the wrong start-up file, and the wrong port. Nothing has changed since August 2025.

Unmaintained · No commits in 12 months.
Coding
Other

Use This MCP client To

Chat with Gemini in a desktop window instead of a browser Attach a tool and watch when Gemini decides to call it Read answers with proper formatting and equations Switch between Gemini models from a settings box Try out a tool I built before wiring it up elsewhere

README

GemCP Chat Logo

GemCP Chat

The GemCP app is a cross-platform desktop application that creates a seamless chat interface for Google's Gemini AI models with extensible capabilities through a Model Context Protocol (MCP) framework.

✨ Features

  • 🤖 Gemini Integration: Seamless chat interface with Google's Gemini models (configurable via Settings).
  • 🔧 Extensible Tools (MCP): Connect external tools and data sources via the Model Context Protocol.
    • Supports Python-based MCP servers (added via file path).
    • Supports command-based MCP servers (e.g., Node.js) defined in a JSON configuration file.
  • 🖥️ Cross-Platform: Runs on macOS and Windows (Electron build).
  • 📊 Tool Status UI: Provides visual feedback when Gemini is calling an MCP tool and whether it succeeded or failed.
  • ⚙️ Model Selection: Choose different Gemini models (e.g., 1.5 Flash, 1.5 Pro, 2.5 Pro Exp) through the Settings dialog.
  • 📝 Markdown & LaTeX Rendering: Displays AI responses with formatting.

Screenshots

Weather Tool Example
Example of the weather tool in action

Chat Interface
Main chat interface with Gemini

Calculator Tool Example
Using the calculator tool with Gemini

Overview

This repository contains both a Python backend and an Electron-based desktop application for interacting with Gemini.

Project Structure

  • mcp-gemini-desktop/: Frontend Electron application
  • python_backend/: Python backend server

Running the Python Backend

Prerequisites

  • Python 3.13+ (as specified in pyproject.toml)
  • uv (Modern Python package installer and resolver)

Setup

  1. Navigate to the Python backend directory:

    cd python_backend

  2. Install uv if you don't have it already:

    pip install uv

  3. Install the required dependencies using uv:

    uv pip install .

  4. Set your Google API key as an environment variable:

    # For Linux/macOS
export GOOGLE_API_KEY=your_api_key_here

# For Windows
set GOOGLE_API_KEY=your_api_key_here

  5. Start the Python backend server:

    python main.py

    The server should start running on http://localhost:5000

Running the Frontend Electron App

Prerequisites

  • Node.js (v16+)
  • npm (Node package manager)

Setup

  1. Navigate to the Electron app directory:

    cd mcp-gemini-desktop

  2. Install the required dependencies:

    npm install

  3. Start the Electron app in development mode:

    npm start

Building the App

To build the Electron app for your platform:

cd mcp-gemini-desktop
npm run build

This will create platform-specific binaries in the dist folder.

Using Pre-built Binaries

If you prefer to use pre-built binaries directly:

  1. Navigate to the mcp-gemini-desktop/dist directory
  2. For macOS: Install the .dmg file
  3. For Windows: Run the .exe installer

Available Binaries

  • macOS (Apple Silicon): GemCP Chat-0.1.0-arm64.dmg
  • Windows: Check the dist folder for .exe files

Example Servers

The repository includes example server implementations in mcp-gemini-desktop/mcp_example_servers/:

  • mcp_server_calc.py: Calculator server example
  • mcp_server_weather.py: Weather information server example

Troubleshooting

  • Ensure the Python backend is running before starting the Electron app
  • Check that your Google API key is correctly set
  • Verify that the required ports are not blocked by a firewall

License

This project is licensed under the MIT License - see the LICENSE file for details.

Contributing

Please raise PR for any contributions. Any PR raised will be reviewed and merged with the main branch.

gemini-desktop FAQ

Does it actually work right now?
Not without editing the code. The four models it can choose from are hard-coded, all of them have been retired by Google, and any other name is refused, so messages fail.
Do I need a key?
Yes, a Google Gemini key. Be warned that the instructions tell you to set GOOGLE_API_KEY while the code only ever reads GEMINI_API_KEY, so following the guide leaves it unable to start.
Does it cost money?
The app is free, but Gemini usage bills to your Google account once you pass the free allowance.
Where does my key end up?
If you type it into the settings box it is held in memory only and disappears when you quit, so you re-enter it each time unless you put it in an environment file.
Can I use this to run tools I already have?
In principle yes — it accepts Python tool files and command-launched ones listed in a settings file — but the dead model list stops the conversation before any tool is reached.
How hard is setup?
Hard. You install Python 3.13 and the uv tool, start a background service by hand in a terminal, then launch the desktop app separately, and the published steps point at the wrong file and port.
Is it maintained?
No code has changed since August 2025, which is why the model list has been overtaken by Google's retirements.
Which computers does it run on?
Mac and Windows, with ready-made installers included in the project.