Image Generation MCP Server

An MCP (Model Context Protocol) server for image generation and editing using multiple providers (Gemini, Replicate, Hugging Face), plus local image manipulation tools.

Installation

Option 1: Global Installation via npm

npm install -g mcp-image-gen

Option 2: Local Installation

npm install mcp-image-gen

Configuration

You can configure the provider through environment variables. The server will automatically select the provider based on configuration or API key availability.

Supported Providers

1. Google Gemini (Default)

• Model: gemini-3-pro-image-preview
• Environment Variable: GEMINI_API_KEY
• Cost: Free (currently in preview)

2. Replicate

• Generation: sdxl-lightning (Fast and low cost)
• Editing: instruct-pix2pix
• Environment Variable: REPLICATE_API_TOKEN
• Explicit Configuration: IMAGE_GENERATION_PROVIDER=replicate

3. Hugging Face

• Generation: stable-diffusion-xl-base-1.0
• Editing: Not supported in current version
• Environment Variable: HUGGING_FACE_TOKEN
• Explicit Configuration: IMAGE_GENERATION_PROVIDER=huggingface

.env Example

# Gemini
GEMINI_API_KEY=your-gemini-api-key

# Replicate
REPLICATE_API_TOKEN=your-replicate-api-token

# Hugging Face
HUGGING_FACE_TOKEN=your-hugging-face-token

# Force provider selection (optional)
IMAGE_GENERATION_PROVIDER=replicate

MCP Client Configuration

Claude Desktop / Amp

Add the following to your configuration file (e.g., claude_desktop_config.json):

{
  "mcpServers": {
    "image-generation": {
      "command": "npx",
      "args": ["-y", "mcp-image-gen"],
      "env": {
        "GEMINI_API_KEY": "your-gemini-api-key",
        "REPLICATE_API_TOKEN": "optional-replicate-token",
        "HUGGING_FACE_TOKEN": "optional-hf-token"
      }
    }
  }
}

Manual Execution (Testing)

To run the server manually:

npx -y mcp-image-gen

Option 2: Using Node directly

node path/to/mcp-server.js

Option 3: Using global installation

If installed globally with npm install -g mcp-image-gen:

image-generation-mcp-server

Environment Variables: Make sure to pass the required environment variables (GEMINI_API_KEY, REPLICATE_API_TOKEN, HUGGING_FACE_TOKEN) in your client configuration.

Available Tools

The parameters available for each tool depend on the configured provider.

generate_image_from_text

Generates a new image from a text description.

Base Parameters:

• prompt: Detailed description of the image.
• output_path (optional): Path where the generated image will be saved.

Gemini Provider Extra Parameters: When using Google Gemini, these additional parameters are available:

• aspectRatio: Aspect ratio of the image. Supported values: 1:1 (default), 3:4, 4:3, 9:16, 16:9.
• resolution: Resolution/Size of the image. Supported values: 1K (default), 2K, 4K.
• numberOfImages: Number of images to generate (candidate count).

Note: Gemini generates images in PNG format.

edit_image

Edits an existing image based on instructions. Note: This tool only modifies the image content (visuals); it does NOT change the image format or dimensions.

Base Parameters:

• image_path: Path to the original image.
• prompt: Editing instructions.
• output_path (optional): Path where the result will be saved.

Gemini Provider Extra Parameters:

• aspectRatio, resolution, numberOfImages.

remove_background

Removes the background from an image.

Parameters:

• image_path: Path to the image file.
• output_path (optional): Path where the transparent image will be saved.

convert_image_format

Converts an image to a different format (e.g., PNG, JPEG, WEBP).

Parameters:

• source_path: Path to the source image.
• format: Target format (png, jpeg, jpg, webp, gif, tiff, avif).
• output_path (optional): Path where the converted image will be saved.

resize_image

Resizes an image to specific dimensions.

Parameters:

• source_path: Path to the source image.
• width: Target width in pixels.
• height (optional): Target height in pixels.
• fit (optional): How the image should be resized to fit the dimensions (cover, contain, fill, inside, outside). Default is cover.
• output_path (optional): Path where the resized image will be saved.

get_image_info

Get metadata about an image (dimensions, format, etc.).

Parameters:

• image_path: Path to the image file.

Requirements

• Node.js 18+
• API key from at least one of the supported providers.