Openai Gpt Image MCP

What SureScaleAI/openai-gpt-image-mcp MCP server does

SureScaleAI/openai-gpt-image-mcp MCP server connects MCP-compatible clients to OpenAI image APIs. It supports image generation from text prompts and image editing tasks such as inpainting, outpainting, and compositing. The repository describes support for OpenAI GPT-4o and gpt-image-1 image workflows.

The server exposes two tools:

• create-image generates one or more images from a prompt and accepts options including size, quality, and background.
• edit-image changes or extends an existing image using a prompt. An optional mask controls where edits are applied.

Input images and masks can be supplied as file paths or base64 data. Generated results may be returned as base64 data or saved to disk.

How it works

The project is a TypeScript Node.js application. After building the source, an MCP client starts dist/index.js as a local process and passes credentials through its environment. The server then calls the configured OpenAI image API and returns the resulting image data or output paths through MCP.

create-image can request up to 10 images in one operation by setting n. For larger results, the server avoids returning an oversized tool response. If the combined image data exceeds the 1MB response limit used by clients such as Claude Desktop, it automatically writes the images to disk and returns their paths instead.

When no output path is supplied, these larger results are saved under /tmp, unless MCP_HF_WORK_DIR changes the working directory. File paths used for inputs must be absolute, and output directories must be writable.

Setup and configuration

Install and build the repository with the documented Yarn workflow:

git clone https://github.com/SureScaleAI/openai-gpt-image-mcp.git
cd openai-gpt-image-mcp
yarn install
yarn build

Configure the MCP client to run Node with the built dist/index.js file. The standard setup uses OPENAI_API_KEY. The README also documents Azure deployments using AZURE_OPENAI_API_KEY, AZURE_OPENAI_ENDPOINT, and OPENAI_API_VERSION instead. An environment file can be passed with the --env-file argument.

SureScaleAI/openai-gpt-image-mcp MCP server requires an OpenAI organization with image API access. The documentation notes that verified organizations may need 15–20 minutes before image access becomes active.

Tools and capabilities

SureScaleAI/openai-gpt-image-mcp MCP server is suited to clients including Claude Desktop, Cursor, VSCode, and Windsurf. Its documented capabilities are:

• Text-to-image generation
• Prompt-driven image editing
• Mask-based edit control
• Multiple image generation, up to 10 per request
• File-path or base64 image input
• File or base64 result output
• Automatic disk output for large responses

Limitations and notes

Base64 responses can exceed the 1MB limit imposed by MCP clients. File output is preferable for high-resolution or multi-image requests. Check image extensions and formats when file-type errors occur, and use absolute paths on Unix-like systems or Windows drive-letter paths on Windows. The project is released under the MIT license.