Open-AutoGLM is an open-source mobile-agent framework built around AutoGLM for controlling smartphones through natural-language instructions. The system uses multimodal models to understand what is displayed on a phone screen, plans the necessary actions, and operates Android devices through ADB to complete user tasks.

A user can describe a goal such as opening an application and searching for particular content, and the agent translates that request into a sequence of visual understanding and device-control steps. The project therefore combines vision-language reasoning, planning, and low-level mobile automation rather than depending on application-specific APIs.

Open-AutoGLM is useful for researchers and developers exploring general-purpose smartphone agents, accessibility-style automation, and natural-language control of mobile applications. It provides a practical foundation for studying how multimodal agents can perceive and operate graphical interfaces that were originally designed only for human touch interaction.