mcp-image

The MCP Image Server is a TypeScript-based service that generates custom images and icons using OpenAI's GPT Image 1 integration. It can produce PNG, SVG, and ICO outputs, with configurable sizes (1024x1024, 1536x1024, or 1024x1536) and backgrounds (transparent or opaque). The server exposes tools such as generate-image for creating images from prompts, and generate-favicon for producing favicon-style icons. It is designed to be run on-demand via MCP workflows and can be integrated with MCP clients, agents, and Vibe Coding flows. To use it, run the server via npx and provide prompts and desired formats; the server will return the generated assets saved to disk or delivered directly as part of the MCP response. The environment variable OPENAI_API_KEY is required for authentication with GPT Image 1. The repository also exposes resources for tool usage examples and supported formats to help you build MCP-driven pipelines.

Prerequisites:

• Node.js and npm installed on your system
• Access to an OpenAI API key

Installation steps:

1. Install or run via npx (no prior installation required):

   • npm install -g npm
   • npx -y @ricardopera/mcp-image-server@latest

2. If you prefer local installation from source:

   • git clone https://github.com/ricardopera/mcp-image-server.git
   • cd mcp-image-server
   • npm install
   • npm run build

3. Set the required environment variable for OpenAI:

   • On Windows (PowerShell): $env:OPENAI_API_KEY="your-key-here"
   • On Linux/macOS: export OPENAI_API_KEY="your-key-here"

4. Run the server via npx (as shown in the example):

   • npx mcp-image-server

5. Verify installation by triggering a sample tool (e.g., generate-image) through an MCP client or the provided test script.

Tips and known considerations:

• The server supports PNG, SVG, and ICO formats. SVG generation is noted as temporarily unavailable in the docs; use PNG or ICO until re-enabled.
• Ensure OPENAI_API_KEY is valid and has access to GPT Image 1 features.
• If you encounter OPENAI_API_KEY not defined errors, double-check environment variable export in your shell or session.
• The MCP integration expects an stdio-based command with appropriate args; adjust the server entry in your MCP client configuration accordingly.
• The project is cross-platform; if you experience permission issues on Windows, try running the terminal as administrator.
• You can customize image sizes, backgrounds, and quality via the provided tool parameters when making requests through MCP clients.
• This server is designed to integrate with MCP.so workflows and can be added to the sample_config.json used by MCP clients.