Imagen3 MCP

What hamflx/imagen3-mcp MCP server does

The hamflx/imagen3-mcp MCP server connects an MCP client to Google Imagen 3.0 for image generation. An agent can provide a descriptive prompt, and the server uses the configured Google API access to generate an image. The examples in the project cover a photorealistic running dog and a stylized, technology-themed apple, showing that prompts can describe subjects, lighting, camera characteristics, and visual style.

This server is intended for clients that can launch an MCP process locally. The repository documents setup for Cherry Studio and Cursor, and distributes executable files through GitHub Releases. It is not described as an image editing, image analysis, or general-purpose media management service.

How it works

The executable starts an MCP service and uses GEMINI_API_KEY to authenticate requests to Google's Imagen 3.0 service. Generated image resources are exposed through URLs. By default, the server listens on 127.0.0.1 and uses port 9981; these values can be changed for different deployment layouts.

IMAGE_RESOURCE_SERVER_ADDR controls the address placed in image URLs, which is useful when the process runs in a container or on another machine. SERVER_LISTEN_ADDR controls the network address where the local service listens. These settings are separate: one determines the bind address, while the other determines how clients reach image resources.

Setup and configuration

To use the hamflx/imagen3-mcp MCP server, download the latest executable from the project's GitHub Releases page and place it on the host where the MCP client runs. Configure the executable path as the client's command, then provide a valid Google Gemini API key through GEMINI_API_KEY.

Cursor configuration uses an mcpServers entry with the executable path in command and the environment variables under env. Cherry Studio uses equivalent command and environment-variable settings. The repository does not provide a package-manager installation command; its documented installation path is a release executable.

Optional settings include:

• BASE_URL for a proxy or alternate API base URL.
• SERVER_LISTEN_ADDR for the address the service binds to; the default is 127.0.0.1.
• SERVER_PORT for the listening port and image URL port; the default is 9981.
• IMAGE_RESOURCE_SERVER_ADDR for the host address inserted into generated image URLs; the default is 127.0.0.1.

A proxy can help with connectivity to Google's endpoint in some network environments, but the README notes that it does not remove Google's IP restrictions.

Tools and capabilities

The documented capability is text-to-image generation through Imagen 3.0 over MCP. Prompts may request photographic or artistic results and can include scene details such as lighting, lens characteristics, subject matter, and desired visual quality. The provided material does not name individual MCP tools or expose additional operations beyond image generation.

Limitations and notes

A Google Gemini API key is mandatory, so users must supply their own credentials and account access to Google's service. Network access to Google may still be required even when BASE_URL points to a proxy. The material does not specify image dimensions, output formats, generation limits, model-selection options, or error-handling behavior.

The project is licensed under MIT. The README confirms Cursor and Cherry Studio setup, but does not document compatibility with other MCP clients.