y. * Replace `sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx` with your actual **OpenAI API Key**. * Ensure `disabled` is `false` to enable the server. The MCP client should automatically detect the changes and connect to the server. ## Tool Usage (`create_image`) Once installed and configured, the server provides the `create_image` tool. ### Accessing Documentation Resources The server exposes its documentation as MCP resources which can be accessed by LLMs before crafting prompts: ```xml image-mcp-server docs/prompt-recipes image-mcp-server docs/readme ``` This allows LLMs to read the prompt recipes documentation before crafting image prompts, leading to better results. * **Description:** Generates a new image based on a text prompt, optionally applying branding or using a style definition JSON, and saves it to a target project's public folder. Can also edit or combine existing images by providing reference images. * **Detailed Usage & Parameters:** For comprehensive details on parameters (`prompt`, `filename`, `targetProjectDir`, `brandSignature`, `styleDefinitionJSON`, `model`, `size`, `quality`, `background`, `referenceImagePaths`, etc.) and guidance on how an LLM should use this tool effectively, please refer to the **detailed documentation comment block at the top of the `src/index.ts` file** in this repository. * **Background Options:** The `background` parameter accepts either 'transparent' or 'opaque' (default) values. Transparent backgrounds are useful for logos, icons, and overlays, and work best with the `gpt-image-1` model. Transparent backgrounds require PNG or WebP format. * **Image Editing:** When `referenceImagePaths` is provided (an array of paths relative to the target project's public folder), the tool will use OpenAI's image editing capabilities to modify or combine those images based on the prompt. For image editing operations, the model is automatically set to `gpt-image-1`. * **Style Definition:** The optional `styleDefinitionJSON` parameter allows you to define a consistent visual style across multiple images. This is especially useful for creating themed image collections with a unified aesthetic. The JSON structure is passed directly to the image model within the prompt, allowing for flexible and expressive style definitions. * **Output:** Returns a JSON object containing `{ ok: true, path: "relative/path/to/image.png", operation: "edit"|"generate", ... }` on success, where `path` is relative to the `public` folder of the `targetProjectDir` provided in the request. **Example `curl` Test (after starting the server manually for testing):** ```bash # First, build and start the server from its directory: # npm run build && node dist/index.js # In another terminal: curl -X POST http://localhost:5050/createImage \ -H 'Content-Type: application/json' \ -d '{ "prompt": "A cute otter mascot", "filename": "otter-mascot.png", "targetProjectDir": "/path/to/your/target/project" }' # Example with background parameter for transparent PNG: curl -X POST http://localhost:5050/createImage \ -H 'Content-Type: application/json' \ -d '{ "prompt": "A professional logo of a cloud with lightning bolt", "filename": "cloud-logo.png", "model": "gpt-image-1", "background": "transparent", "targetProjectDir": "/path/to/your/target/project" }' # Example with image editing and transparent background: curl -X POST http://localhost:5050/createImage \ -H 'Content-Type: application/json' \ -d '{ "prompt": "Add a glowing blue aura effect around the object", "filename": "enhanced-icon.png", "model": "gpt-image-1", "background": "transparent", "referenceImagePaths": ["existing-icon.png"], "targetProjectDir":

Image Generation MCP Server

Comments