Gemini Skill

What it does

Gemini Skill provides a comprehensive bridge to the Gemini web interface using Chrome DevTools Protocol (CDP). It allows AI agents to bypass API limitations by directly controlling a browser instance to perform complex tasks like AI image generation and multimodal conversations.

Tools

• gemini_generate_image: Full end-to-end image generation process including high-res downloads.
• gemini_send_message: Sends text prompts and retrieves multi-turn conversation responses.
• gemini_upload_images: Uploads reference images for image-to-image or vision-based tasks.
• gemini_extract_image: Extracts specific images from a session as base64 or full-size files.
• gemini_switch_model: Toggles between different Gemini models (e.g., Pro, Flash, Think).
• gemini_new_chat: Initializes a fresh conversation session.

Installation

To use this server, you must have a Chrome/Edge browser installed and be logged into your Google account.

{
  "mcpServers": {
    "gemini": {
      "command": "node",
      "args": ["/absolute/path/to/gemini-skill/src/mcp-server.js"]
    }
  }
}

Supported hosts

Confirmed for use with Claude Desktop and OpenClaw agents.