wtf is translategemmamcp?

TranslateGemmaMCP lets you run text translation on your own computer using a local AI model called TranslateGemma, instead of sending text to a cloud translation service. It packages this local model behind a Model Context Protocol server, a standard way for AI tools and agents to call external capabilities, so any MCP-compatible client can ask it to translate text. The project has two parts. A simple web page you open directly in a browser gives you a basic interface for translating text by pointing it at your running TranslateGemma server. The main piece is a Python server that exposes a single translate tool over MCP, which takes the text to translate along with source and target language codes, using standard codes like en or de-DE, and returns the translated result. Because everything runs locally, no text is sent to an outside company, which the README highlights as a privacy benefit. To use it, you first need TranslateGemma itself running through llama.cpp, a tool for running AI models on your own hardware, exposed as an OpenAI-compatible API endpoint. You then install a handful of Python packages and start the MCP server, which listens on a configurable host and port and forwards translation requests to that local model. The server includes automatic retries and rate limiting to keep things stable under repeated use. This project is aimed at developers who already run local AI models and want to plug real translation ability into their own tools, chatbots, or AI agents without relying on an external translation API. It is released under the MIT license.