Repolode
Submit
WJZ-P

WJZ-P/gemini-skill

An MCP server and agent skill that drives the Gemini web app through Chrome DevTools Protocol for image generation, chat, and image extraction.

View on GitHub
Stars
832
Forks
122
Open issues
10
Last commit
last month

What is it?

Gemini Skill connects to a locally running Chrome or Edge instance over CDP and operates the Gemini web interface on the user's behalf. It exposes these actions as MCP tools, so any MCP-compatible client or agent, including OpenClaw, can send prompts, generate images, upload reference images, switch models, and read replies without a Gemini API key.

A background daemon keeps the browser session alive between calls and shuts it down after 30 minutes of inactivity. The browser uses puppeteer-extra stealth plugins, and downloaded images can have the Gemini watermark removed automatically. The project can also be imported as a library or run from the command line.

Key features

  • 01Generates images from prompts and downloads full-size originals via CDP
  • 02Uploads reference images for image-to-image generation
  • 03Starts new, temporary, or historical chats and switches models
  • 04Extracts images from a conversation as base64 or saved files
  • 05Removes the Gemini watermark from downloaded images
  • 06Exposes all actions as MCP tools for any compatible client

Use cases

  • Letting an AI agent generate illustrations or memes as part of a workflow
  • Batch-generating images through Gemini without manual browser steps
  • Using Gemini's image model from tools that only support MCP

Quick start

  1. 1

    Clone the repository

    git clone https://github.com/WJZ-P/gemini-skill.git
  2. 2

    Enter the project folder

    cd gemini-skill
  3. 3

    Install dependencies

    npm install
  4. 4

    Start the MCP server, then add it to your MCP client config (requires Node.js 18+ and Chrome/Edge with a Google login)

    npm run mcp

Things to know

  • Requires a locally installed Chrome, Edge, or Chromium and a logged-in Google account to use Gemini.
  • Image generation is slow, typically 60 to 120 seconds per request.
  • Drives the Gemini web UI via browser automation, so it can break when Google changes the page.

In collections

MCP serversAI image generators & Stable Diffusion tools

Topics

#mcp#browser-automation#gemini#image-generation#cdp#puppeteer#openclaw#automation#drawing#mcp-client

FAQ

What is gemini-skill?
An MCP server and agent skill that drives the Gemini web app through Chrome DevTools Protocol for image generation, chat, and image extraction. Gemini Skill connects to a locally running Chrome or Edge instance over CDP and operates the Gemini web interface on the user's behalf. It exposes these actions as MCP tools, so any MCP-compatible client or agent, including OpenClaw, can send prompts, generate images, upload reference images, switch models, and read replies without a Gemini API key.
Is gemini-skill free to use? What license does it have?
gemini-skill is open source under the MIT license. Check the license terms before using it in commercial products.
Is gemini-skill still maintained?
Updated within 6 months. The last commit to gemini-skill was on Sep 18, 2026.
What can I use gemini-skill for?
Letting an AI agent generate illustrations or memes as part of a workflow Batch-generating images through Gemini without manual browser steps Using Gemini's image model from tools that only support MCP
How do I install gemini-skill?
Clone the repository: git clone https://github.com/WJZ-P/gemini-skill.git Enter the project folder: cd gemini-skill Install dependencies: npm install Start the MCP server, then add it to your MCP client config (requires Node.js 18+ and Chrome/Edge with a Google login): npm run mcp
Where can I get gemini-skill?
The source code and installation instructions are on GitHub: https://github.com/WJZ-P/gemini-skill

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects