MCP Tools Overview

GPT Image 1 MCP provides three powerful tools for image generation and editing through the Model Context Protocol, powered by OpenAI's gpt-image-1 model.

Available Tools

generate-image

Generate images using OpenAI's gpt-image-1 model with advanced text rendering and superior instruction following. Note: Prompts must be in English for optimal results.

Parameters:

• prompt (string, required): Image description
• aspect_ratio (string): "square", "landscape", or "portrait"
• quality (string): "standard" or "hd"
• style (string): "vivid" or "natural"
• output_directory (string): Directory to save the image
• filename (string): Custom filename
• save_to_file (boolean): Whether to save locally

Returns:

• Image URL
• Local file path (if saved)
• Generation metadata

edit-image

Edit existing images with AI-powered modifications using gpt-image-1 model. Supports inpainting, outpainting, style transfer, and variations with native transparency. Note: Edit prompts must be in English for optimal results.

Parameters:

• source_image (string, required): Image URL or base64 encoded image
• edit_prompt (string, required): Description of desired changes
• edit_type (string, required): Type of edit ("inpaint", "outpaint", "variation", "style_transfer", etc.)
• strength (number): Edit strength (0.0 = minimal, 1.0 = maximum)

Returns:

• Edited image file path
• Metadata about the edited image

batch-edit

Apply the same edit to multiple images efficiently using gpt-image-1 model. Supports parallel processing for improved performance. Note: Edit prompts must be in English for optimal results.

Parameters:

• image_urls (string[], required): Array of image URLs to edit
• edit_prompt (string, required): Edit description to apply to all images
• edit_type (string, required): Type of edit to apply
• batch_settings (object): Configuration for batch processing

Returns:

• Array of edited image file paths
• Processing status for each image

Image Analysis

For image analysis capabilities, LLM clients can directly read image files using the file paths returned by the generation and editing tools. This provides better separation of concerns and allows the LLM to use its native vision capabilities.

Tool Response Format

All tools follow a consistent response format:

{
  "success": boolean,
  "data": {
    // Tool-specific data
  },
  "error": {
    "message": string,
    "code": string
  }
}

Error Handling

Common error codes:

• INVALID_PARAMS: Invalid or missing parameters
• API_ERROR: OpenAI API error
• RATE_LIMIT: Rate limit exceeded
• AUTH_ERROR: Authentication failed

Usage Examples

See the Examples section for detailed usage patterns.