TuriX – Desktop Actions, Driven by AI

TuriX is an open‑source Model Context Protocol (MCP) server that turns any advanced AI assistant into a fully autonomous desktop operator. By exposing a rich set of resources, tools and prompts, it allows models such as Claude to issue high‑level instructions that are translated into real mouse clicks, keyboard strokes and application API calls. The result is a seamless hand‑off from conversational intent to tangible computer actions, all while keeping the system 100 % free for personal and research use.

The server solves a core pain point in AI‑powered automation: the gap between abstract natural‑language commands and the concrete, often proprietary interfaces of desktop software. Traditional bots rely on hard‑coded UI maps or vendor APIs, which limits portability and requires constant maintenance. TuriX eliminates these constraints by adopting a no‑app‑specific‑API philosophy—if a human can click a button, the server can too. This makes it possible to automate complex workflows across a wide range of applications (WhatsApp, Excel, Outlook, in‑house tools) without writing custom adapters.

Key capabilities are packaged into a clean MCP interface. Developers can swap the underlying brain—the vision‑language policy that interprets screenshots and text—simply by editing . The default model already outperforms earlier open‑source agents, achieving over 68 % success on the OSWorld test set and maintaining fast response times on macOS. The server also provides sampling and prompting hooks, allowing fine‑grained control over how the AI generates actions and handles uncertainty.

Real‑world scenarios that benefit from TuriX include automated travel booking, data entry across spreadsheets and word processors, social media interaction, and sophisticated cross‑application tasks such as extracting data from a spreadsheet, generating charts in Numbers, inserting them into PowerPoint slides, and emailing the final deck—all triggered by a single natural‑language request from Claude. The MCP integration means any agent that speaks the protocol—Claude for Desktop, or custom agents built on LlamaIndex or LangChain—can immediately gain desktop control without additional plumbing.

For developers, TuriX offers a plug‑and‑play entry point into desktop automation. By exposing a uniform MCP endpoint, it can be dropped into existing AI workflows: the model receives a prompt, replies with an action sequence, and the server translates that into real UI interactions. This tight coupling eliminates latency, reduces boilerplate code, and provides a single source of truth for permission handling (macOS accessibility, Safari automation). The result is a powerful, extensible platform that scales from personal productivity hacks to enterprise‑grade automation pipelines.