Overview
What is Gemini Imagen 3.0 MCP Server?
A professional Model Context Protocol (MCP) server implementation that harnesses Google’s Imagen 3.0 model through the Gemini API for high-quality image generation. Built with TypeScript, it is designed for seamless integration with Claude Desktop and other MCP‑compatible hosts.
How to use Gemini Imagen 3.0 MCP Server?
Clone the repository, install dependencies (npm install), build the TypeScript code (npm run build), then configure Claude Desktop by adding the server to claude_desktop_config.json with the correct path and your GEMINI_API_KEY. Use the generate_images tool to create images and the create_image_html tool to generate HTML preview tags.
Key features of Gemini Imagen 3.0 MCP Server
- Leverages Google’s Imagen 3.0 model via the Gemini API
- Generates up to 4 high‑quality images per request
- Automatic file management with intelligent naming
- HTML preview generation with file:// protocol support
- Built on MCP protocol for AI agent compatibility
- TypeScript implementation with robust error handling
Use cases of Gemini Imagen 3.0 MCP Server
- AI‑powered image generation directly from Claude Desktop or other MCP hosts
- Batch creation of up to four images from a single text prompt
- Local preview of generated images using automatically created HTML tags
- Integration into automated workflows that require image generation with file management
FAQ from Gemini Imagen 3.0 MCP Server
What tools does the server provide?
It provides two tools: generate_images (prompt required, optional numberOfImages 1‑4) and create_image_html (imagePaths required, optional width/height).
What are the prerequisites to run the server?
You need Node.js 18 or higher, a Google Gemini API key, and an MCP‑compatible host such as Claude Desktop.
How are generated images stored?
Images are saved automatically to G:\image-gen3-google-mcp-server\images with filenames following the pattern {sanitized-prompt}-{timestamp}-{index}.png.
What error codes does the server use?
It implements two error codes: tool_not_found (1) when a requested tool is unavailable, and execution_error (2) when image generation or HTML creation fails.
Can I generate more than four images per request?
No, the numberOfImages parameter accepts a value between 1 and 4 (default 1).
