UFO²

Open-source (MIT) Desktop AgentOS to automate multi-app Windows workflows. Uses hybrid GUI+API control, RAG, & multi-agent framework for robust automation.

UFO² is an open-source Desktop AgentOS designed to automate complex, multi-application workflows on Windows. It translates natural-language requests into reliable automation by combining GUI and API controls, supported by a multi-agent framework. The system features deep OS integration, hybrid control detection, and a speculative executor to reduce latency. Key innovations include a knowledge substrate for continuous learning, a sandboxed virtual desktop (coming soon), and robust benchmarks for real-world tasks. UFO² is ideal for users seeking efficient, scalable automation without disrupting their primary workspace. The tool requires Python 3.10+ on Windows 10+ and supports various LLMs, including OpenAI and Azure. Its modular design allows customization through RAG-enhanced knowledge retrieval. UFO² represents a significant leap in Windows automation, balancing flexibility with performance.