MobileAgent

Why we included this project

Building a GUI agent that can actually operate real software means solving two problems at once: reading what is on screen and deciding what to do next. The Mobile-Agent family tackles both by pairing screen-reading multimodal models with a decision layer that turns visual understanding into concrete actions on mobile apps and desktop programs. The repo also ships open models like GUI-Owl and the newer ToolCUA, along with training data and evaluation resources, so teams can study how a production-grade GUI agent is architected end to end instead of just calling a thin wrapper. That makes it a practical starting point for researchers and engineers moving from single-app demos toward agents that reason about which tool fits a task. Online demos and hosted APIs are linked in the README, so you can try it before committing to a deployment.