Gemini MCP Desktop Client

What it does

Gemini MCP Desktop Client is an Electron-based desktop application that brings MCP server connectivity to macOS and Windows with Google Gemini as the underlying LLM. Rather than requiring a cloud-hosted AI client, you install and run the app locally, point it at any MCP-compatible server, and get full LLM-driven chat with tool invocation powered by Gemini's API.

The app bundles a local Express backend, a React/MUI frontend, and the official @modelcontextprotocol/sdk to handle all MCP server communication — stdio or otherwise. It includes bundled uv/uvx binaries so Python-based MCP servers just work without extra setup.

Tools

This is a desktop MCP client — it does not expose tools itself. Instead it connects to and invokes tools on any MCP server you configure:

• Multi-server support — connect to one or more MCP servers from within the settings UI
• Gemini LLM backend — routes all chat turns and tool calls through @google/generative-ai
• Voice input — microphone capture with audio-to-text (audioToText) pipeline
• File parsing — reads .xlsx, .pptx, .docx via bundled parsers before sending to LLM
• Markdown rendering — full syntax-highlighted code blocks via react-markdown + rehype-highlight
• Drag-and-drop UI — reorder server connections via @hello-pangea/dnd

Installation

// claude_desktop_config.json — not applicable (this IS the client)
// Install directly:
git clone https://github.com/duke7able/gemini-mcp-desktop-client.git
cd gemini-mcp-desktop-client
npm install
npm run create-env   # creates .env from .env.example
npm run dev          # start in development mode

Or download prebuilt installers:

• Windows EXE: Download
• macOS DMG/ZIP: Download

Requires a Google Gemini API key set in .env before first launch.

Supported hosts

This app is itself the MCP host — it connects to MCP servers rather than running inside another client.