Gemini Image Generation MCP

A Model Calling Protocol (MCP) server that allows Claude and other LLMs to generate images using Google's Gemini AI model.

Overview

This MCP server enables Large Language Models (LLMs) like Claude to delegate image generation tasks to Google's Gemini model. It follows the MCP standard to provide a seamless integration experience.

Features

• Generate images from text prompts using Google's Gemini model (gemini-2.0-flash-preview-image-generation)
• Save generated images locally
• Configure generation parameters like temperature, topK, and topP
• Interactive web interface for testing and demonstration
• Docker support for easy deployment
• Simple MCP standard implementation

Prerequisites

• Node.js 18 or higher
• Google API key with access to Gemini API (specifically the image generation preview model)
• For Docker: Docker and Docker Compose

Getting Started

Installation

1. Clone this repository:

git clone https://your-repository-url/gemini-image-gen-mcp.git
cd gemini-image-gen-mcp

2. Install dependencies:

npm install

3. Create your environment configuration:

cp .env.example .env

4. Edit the .env file to add your Google API key:

GEMINI_API_KEY=your_gemini_api_key_here

Running the Server

Option 1: Using Node.js directly

For MCP server only:

node src/mcp-server.js

For web interface only:

node src/web-server.js

Or use the provided script:

# Run both MCP server and web interface
./start-server.sh --both

# Run MCP server only
./start-server.sh --mcp-only

# Run web interface only
./start-server.sh --web-only

# Default (MCP server only)
./start-server.sh

The web interface will be available at http://localhost:3070 (or the port specified in your .env file).

Option 2: Using Docker

docker-compose up -d

Web Interface

The project includes a web interface for testing and demonstrating the image generation capabilities:

• Generator: Create images by entering text prompts and adjusting parameters
• Gallery: View all previously generated images
• About: Information about the project and its features

To access the web interface, navigate to http://localhost:3070 in your browser after starting the web server.

Usage with Claude

API Usage

When making API requests to Claude, include the MCP configuration:

{
  "mcpServers": {
    "openai-image-generation": {
      "command": "node",
      "args": ["/full/path/to/openai-image-gen-mcp/src/mcp-server.js"],
      "env": {
        "OPENAI_API_KEY": "your-openai-api-key-here"
      }
    }
  }
}

Claude Console Usage

1. Go to Developer Settings
2. Navigate to the MCPs section
3. Click "Add MCP"
4. Fill in the details:
   • Name: gemini_image_generation
   • URL: https://your-server-url.com/mcp
   • Authentication (if implemented): Select Bearer Token and enter your token

Gemini Image Generation API

This server uses the Gemini API for image generation. Specifically, it uses the gemini-2.0-flash-preview-image-generation model which is optimized for image generation tasks. The implementation follows Google's official API documentation for properly formatting requests and handling responses.

Key features of the Gemini image generation implementation:

• Uses proper response modality for image generation
• Handles base64-encoded image data from response
• Provides enhanced prompts alongside generated images
• Automatically saves generated images for later use

Configuration Options

The MCP server supports the following configuration options:

License

ISC

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.