This is an AI Agent framework that enables AI to operate computers (such as macOS, Windows, Linux, Android) like humans and automatically complete complex GUI operation tasks. It adopts a 'general body + expert body' combined architecture and supports active hierarchical planning. By integrating large models (LLM) and visual multimodal models, it can understand inputs such as screenshots and interface structures and generate operation instructions to achieve automatic clicking, inputting, window switching, searching and other operations.