img-mcp

📚 View the website

An open-source Model Context Protocol (MCP) server that lets AI agents generate images. It currently supports Google's Nano Banana image-generation models (Gemini Flash Image) through the Gemini API.

Features

• Text-to-image generation via MCP tools
• Image-to-image / reference-image generation by passing base64 images or data URIs
• Transparent background generation
• Local image resizing — scale, crop-fit, and convert formats without an API key
• Local image format conversion — JPEG, PNG, WebP, AVIF, GIF
• Google Nano Banana support (gemini-3.1-flash-image, gemini-3.1-flash-lite-image, gemini-3-pro-image, etc.)
• User-provided API key — you bring your own Google API key
• Zero-install usage with npx img-mcp
• Multiple aspect ratios and image sizes

Installation

Via npx (recommended)

npx -y img-mcp

Via npm

npm install -g img-mcp
img-mcp

Configuration

Set your Google API key as an environment variable:

export GOOGLE_API_KEY="your-google-api-key"

You can get an API key from Google AI Studio.

Image sizes

Supported image_size values are 512, 1K, 2K, and 4K. Note that 2K and 4K require a Nano Banana model that supports high-resolution output, such as gemini-3-pro-image. The default gemini-3.1-flash-image model typically supports 512 and 1K.

Optional environment variables

Optional: background removal support

The remove_background tool relies on @imgly/background-removal-node, which is licensed under AGPL-3.0. To keep img-mcp itself under the permissive MIT license, this dependency is optional and is not installed by default.

To enable background removal, install the peer dependency in your environment:

npm install -g @imgly/background-removal-node

Or, when using npx, install it alongside img-mcp in a local project. If remove_background is called while the package is missing, the tool returns an error with installation instructions.

Usage with MCP hosts

Add the server to your MCP host configuration (e.g., Claude Desktop, Cursor, or Windsurf):

{
  "mcpServers": {
    "img-mcp": {
      "command": "npx",
      "args": ["-y", "img-mcp"],
      "env": {
        "GOOGLE_API_KEY": "your-google-api-key"
      }
    }
  }
}

Available Tools

generate_image

Generates images from a text prompt.

Arguments

Example

{
  "prompt": "A futuristic city at sunset with flying cars",
  "aspect_ratio": "16:9",
  "image_size": "2K",
  "n": 1
}

Example with reference images

{
  "prompt": "Generate a portrait in the same style as the reference photo",
  "aspect_ratio": "1:1",
  "image_size": "1K",
  "images": [
    "data:image/png;base64,iVBORw0KGgo..."
  ]
}

Output behavior

• If output_path is provided, the generated image(s) are written to disk and the tool returns the saved file path(s). When n is greater than 1, additional images are saved with numbered suffixes (e.g., image_1.png, image_2.png).
• If output_path is omitted, the generated image(s) are returned as base64 content.

resize_image

Resizes an image using fast local processing (powered by sharp). Scale by width, height, percentage, or fit mode.

Arguments

At least one of width, height, or scale must be provided.

Examples

Scale to a specific width:

{
  "image": "data:image/png;base64,iVBORw0KGgo...",
  "width": 800
}

Resize to exact dimensions with cover fit:

{
  "image": "data:image/png;base64,iVBORw0KGgo...",
  "width": 800,
  "height": 600,
  "fit": "cover"
}

Scale by percentage and convert to WebP:

{
  "image": "data:image/png;base64,iVBORw0KGgo...",
  "scale": 0.5,
  "format": "webp",
  "quality": 85,
  "output_path": "/tmp/resized.webp"
}

Output behavior

• If output_path is provided, the resized image is written to disk and the tool returns the saved file path.
• If output_path is omitted, the resized image is returned as base64 content.
• When only width or height is provided, the aspect ratio is preserved.
• When format is omitted, the output format matches the input image format.

remove_background

Removes the background from an image and returns a transparent PNG. Requires the optional peer dependency @imgly/background-removal-node.

Arguments

Example

{
  "image": "/path/to/photo.jpg",
  "output_path": "/path/to/photo-transparent.png"
}

Output behavior

• The output is always a transparent PNG.
• If output_path is provided, the extension is forced to .png.

Development

# Install dependencies
npm install

# Run in development mode
npm run dev

# Build
npm run build

# Lint / type check
npm run lint

Testing with the MCP Inspector

npm run build
npx @modelcontextprotocol/inspector dist/index.js

Provider: Google Nano Banana

This server uses the Gemini API interactions endpoint for Nano Banana models. You are responsible for your own API usage and costs.

License

MIT

Contributing

Contributions are welcome! Please open an issue or pull request on GitHub.