gemini

Gemini MCP provides a suite of grounded research and content generation tools integrated into your MCP desktop environment. The server exposes capabilities such as live web grounding for Chat queries, deep research with multi-step iteration, image and SVG generation with linked search results, and media creation including video. You can access these tools from within Claude Desktop using tool calls like gemini:gemini_chat, gemini:gemini_deep_research, gemini:generate_image, gemini:generate_video, gemini:generate_svg, and related editing or analysis tools. Grounding by default means answers pull in current sources from Google, with markdown links to sources, while turning grounding off lets you reason over internal model outputs when needed. The Gemini tools support adjustable thinking levels and per-call parameters to tailor depth and speed.

Prerequisites:

• Node.js and npm installed on your system (Node.js 14+ recommended)
• Access to a terminal or command prompt
• An API key for Gemini (GEMINI_API_KEY) if you intend to use live grounding

Quick start (short setup using npx):

# Install and run via npx (no local install required)
npm --version
node --version
# Start the MCP server via npx (this fetches and runs the package on demand)
npx @houtini/gemini-mcp

If you prefer to install locally and build for development:

git clone https://github.com/houtini-ai/gemini-mcp
cd gemini-mcp
npm install --include=dev
npm run build

Then point Claude Desktop at the local build:

{
  "mcpServers": {
    "gemini": {
      "command": "node",
      "args": ["C:/path/to/gemini-mcp/dist/index.js"],
      "env": {
        "GEMINI_API_KEY": "your-api-key-here"
      }
    }
  }
}

If you only want to test quickly without building, you can also run the package directly from npm and supply the API key as an environment variable:

GEMINI_API_KEY=your-api-key-here npx @houtini/gemini-mcp

Notes and tips:

• Ensure GEMINI_API_KEY is set in your environment to enable live grounding and access to Google search results.
• If you encounter rate limits, consider using a lower grounding level or staggering requests.
• For local builds, the dist/index.js path must be the output of your build step; adjust the CLI path accordingly in your MCP config.
• The Gemini tools support multiple subcommands (e.g., gemini_chat, gemini_deep_research, generate_image, generate_video, generate_svg, edit_image). Refer to the README for exact parameter shapes per tool. If grounding feels too noisy, you can toggle grounding at the tool call level.
• When using image/video/SVG generation, ensure you have any required system dependencies installed (e.g., ffmpeg for video processing) if you plan to generate media locally.