Artin0123 Gemini Image MCP Server

What Artin0123-gemini-image-mcp-server does

Artin0123-gemini-image-mcp-server provides two MCP tools backed by the Gemini API. analyze_image accepts one or more image URLs in a single request and asks Gemini to analyze the referenced content. analyze_youtube_video accepts a YouTube video URL and sends the video for analysis.

The repository describes this as a remote server. Its remote-oriented workflow is intended for URLs rather than local media files; the README points users who need local image and video handling to a separate local version in the repository. The server is suitable for agents that need to inspect externally hosted images or extract information from YouTube content without first downloading videos themselves.

How it works

The MCP client starts the server with Node and communicates with the compiled JavaScript entry point. When an image request arrives, the server downloads the image from its URL and forwards it for Gemini processing. Each image must be no larger than 16 MB. If a file exceeds that threshold, the request returns an error that reports the actual size.

YouTube requests use the supplied video URL. The README states that these videos are streamed directly through the Gemini API, so the server does not download them locally and does not apply a stated file-size limit. Results depend on the selected Gemini model and the request made by the calling agent.

Setup and configuration

Install the dependencies with npm install, then compile the TypeScript source with npm run build. The resulting server entry point is dist/index.js. A typical MCP configuration starts it with node and passes the absolute path to that file as an argument.

A Gemini API key is required for tool calls. Create the key through Google AI Studio and expose it as GEMINI_API_KEY in the server environment. The process may start without a key, which can help automated discovery or scanning, but tool calls will fail with a configuration error until a valid key is supplied.

The default model is models/gemini-flash-lite-latest. Set GEMINI_MODEL to choose another model, or provide modelName through the Smithery or SDK configuration schema. Smithery also provides an installation command for Claude Desktop.

Tools and capabilities

• analyze_image: analyzes one or more images identified by URL. Each image is subject to the 16 MB limit.
• analyze_youtube_video: analyzes a YouTube video from its URL. The documented workflow streams the video through Gemini and does not specify a size limit.

The repository includes tests covering URL forwarding, MIME handling, and configuration fallbacks. These tests describe implementation areas but do not add additional MCP tools or media sources.

Limitations and notes

Artin0123-gemini-image-mcp-server requires a working Gemini API key before either analysis tool can be used. Image inputs must be reachable by URL and remain within the per-image 16 MB limit. The documented image tool does not describe local file input, and the README directs users to a separate local version for local images and videos.

The supported video input described in the repository is a YouTube URL. Other video hosts are not documented. Although YouTube videos have no stated size restriction in this server, requests still depend on the Gemini API and the configured model. The project is licensed under MIT.