MCP 多提供商图像生成服务器

This project provides a server implementation based on the Model Context Protocol (MCP) that supports multiple AI image generation providers including OpenAI, Stability AI, Replicate, and Hugging Face.

🎨 Features

支持的提供商

• OpenAI: DALL-E 2, DALL-E 3, GPT-Image-1
• Stability AI: Stable Diffusion XL, Stable Diffusion v2.1, Stable Diffusion v1.6
• Replicate: Various open-source models including FLUX, SDXL, Playground v2.5
• Hugging Face: Stable Diffusion models via Inference API

核心功能

• 🖼️ text-to-image generation across multiple providers
• ✏️ image-to-image editing (where supported)
• 🔧 Unified API interface for all providers
• ⚙️ Configurable via environment variables and command-line arguments
• 📏 Support for various image sizes, styles, and quality settings
• 💾 Saves generated/edited images to specified paths

Here's an example of generating an image directly in Cursor using the text-to-image tool integrated via MCP:

🚀 Quick Start

You can run the server directly from npm using npx:

# 基础用法（自动检测可用的提供商）
npx multi-provider-imagegen-mcp

# 指定提供商
npx multi-provider-imagegen-mcp --providers openai stability

# 指定模型
npx multi-provider-imagegen-mcp --models dall-e-3 stable-diffusion-xl-1024-v1-0

# 使用短名称
npx mpimg --providers openai stability replicate

📋 Prerequisites

• Node.js (v18 or later recommended)
• pnpm, npm or yarn
• At least one provider API key:
  • OpenAI API key from OpenAI Platform
  • Stability AI API key from Stability AI Platform
  • Replicate API token from Replicate
  • Hugging Face token from Hugging Face

🔧 Integration with Cursor

Integrate this multi-provider server with Cursor to access various image generation models:

1. Open Cursor Settings:
   • Go to File > Preferences > Cursor Settings (or use Ctrl+, / Cmd+,)
2. Navigate to MCP Settings:
   • Search for "MCP" and find "Model Context Protocol: Custom Servers"
3. Add Custom Server:
   • Click "Edit in settings.json" and add configurations:

🎯 Example Configurations

Multi-Provider Setup (Recommended)

"mcpServers": {
  "multi-image-generator": {
    "command": "npx multi-provider-imagegen-mcp --providers openai stability replicate",
    "env": {
      "OPENAI_API_KEY": "your_openai_key",
      "STABILITY_API_KEY": "your_stability_key",
      "REPLICATE_API_TOKEN": "your_replicate_token"
    }
  }
}

Single Provider Setup

"mcpServers": {
  "openai-image-generator": {
    "command": "npx mpimg --providers openai --models dall-e-3",
    "env": {
      "OPENAI_API_KEY": "your_openai_api_key"
    }
  }
}

Advanced Configuration

"mcpServers": {
  "custom-image-generator": {
    "command": "npx multi-provider-imagegen-mcp --providers openai stability --models dall-e-3 stable-diffusion-xl-1024-v1-0",
    "env": {
      "OPENAI_API_KEY": "your_openai_key",
      "STABILITY_API_KEY": "your_stability_key"
    }
  }
}

After saving, the multi-provider image generation tools will be available in Cursor's MCP tool selection.

⚙️ Setup

1. Clone the repository:

   git clone https://github.com/spartanz51/imagegen-mcp.git
   cd imagegen-mcp

2. Install dependencies:

   pnpm install
   # or
   npm install
   # or
   yarn install

3. Configure Environment Variables: Create a .env file in the project root:

   cp .env.example .env

   Edit the .env file and add your API keys (only add the ones you plan to use):

   # OpenAI (for DALL-E models)
   OPENAI_API_KEY=your_openai_api_key_here
       
   # Stability AI (for Stable Diffusion models)
   STABILITY_API_KEY=your_stability_api_key_here
       
   # Replicate (for various open-source models)
   REPLICATE_API_TOKEN=your_replicate_api_token_here
       
   # Hugging Face (for Inference API models)
   HUGGINGFACE_API_KEY=your_huggingface_api_key_here

🔨 Building

To build the TypeScript code into JavaScript:

pnpm run build
# or
npm run build
# or 
yarn build

This will compile the code into the dist directory.

🚀 Running the Server

Development Mode

pnpm run dev
# or
npx tsx src/index.ts [options]

Production Mode

node dist/index.js [options]

Command Line Options

--providers <provider1> <provider2> ...

Specify which providers to enable:

# Enable specific providers
node dist/index.js --providers openai stability

# Enable all available providers (with API keys)
node dist/index.js --providers openai stability replicate huggingface

--models <model1> <model2> ...

Specify which models to allow (applies to all enabled providers):

# Specific models
node dist/index.js --models dall-e-3 stable-diffusion-xl-1024-v1-0

# Combined with providers
node dist/index.js --providers openai stability --models dall-e-3

Auto-Detection

If no options are provided, the server automatically detects available providers based on environment variables:

node dist/index.js

The server will start and listen for MCP requests via standard input/output (StdioServerTransport).

🛠️ MCP Tools

The server exposes the following unified MCP tools that work across all providers:

text-to-image

Generates an image based on a text prompt using any supported provider.

Parameters:

• text (string, required): The prompt to generate an image from
• outputPath (string, required): Absolute path where the output file should be saved
• model (enum, optional): The model to use (format: provider/model or model). Auto-detects provider if not specified
• provider (enum, optional): The provider to use (openai, stability, replicate, huggingface). Auto-detected from model if not specified
• size (enum, optional): Size of the generated image (e.g., 1024x1024, 1792x1024). Defaults to 1024x1024
• style (enum, optional): Style of the image (vivid or natural). Supported by certain models like DALL-E 3
• output_format (enum, optional): Format (png, jpeg, webp). Defaults to png
• output_compression (number, optional): Compression level (0-100). Defaults to 100
• moderation (enum, optional): Moderation level (low, auto). Defaults to low (OpenAI specific)
• background (enum, optional): Background (transparent, opaque, auto). Defaults to auto
• quality (enum, optional): Quality (standard, auto, high, medium, low). Defaults to auto
• n (number, optional): Number of images to generate. Defaults to 1

Example Usage:

{
  "text": "A beautiful landscape with mountains and a lake",
  "outputPath": "/tmp/landscape.png",
  "model": "openai/dall-e-3",
  "size": "1024x1024",
  "style": "vivid"
}

image-to-image

Edits an existing image based on a text prompt and optional mask.

Parameters:

• images (array, required): An array of file paths to local images
• prompt (string, required): A text description of the desired edits
• outputPath (string, required): Absolute path where the output file should be saved
• mask (string, optional): File path of mask image (PNG). Transparent areas indicate where the image should be edited
• model (enum, optional): The model to use (format: provider/model or model). Auto-detects provider
• provider (enum, optional): The provider to use. Auto-detected from model if not specified
• size (enum, optional): Size of the generated image. Defaults to 1024x1024
• output_format (enum, optional): Format (png, jpeg, webp). Defaults to png
• output_compression (number, optional): Compression level (0-100). Defaults to 100
• quality (enum, optional): Quality settings. Defaults to auto
• n (number, optional): Number of images to generate. Defaults to 1

Provider Support:

• OpenAI: Supports DALL-E 2 and GPT-Image-1 for editing
• Stability AI: Supports image-to-image with Stable Diffusion models
• Replicate: Supports various image editing models
• Hugging Face: Limited support (falls back to text-to-image for some models)

Returns:

Both tools return:

• content: An array containing a text object with the path to the saved image file

Development

• Linting: npm run lint or yarn lint
• Formatting: npm run format or yarn format (if configured in package.json)

Contributing

Pull Requests (PRs) are welcome! Please feel free to submit improvements or bug fixes.