TuriX is a computer-use agent framework that allows AI models to take real-world actions directly on a desktop environment. It uses vision-language models (Qwen3-VL) to perceive screen state and execute GUI automation tasks. The system achieves 80% success on an OSWorld-style Mac benchmark and supports customization via config files. It integrates with Model Context Protocol (MCP) and offers skills via OpenClaw/ClawHub for extended capabilities.