roubao

Why we included this project

Roubao is a native Android app, written in Kotlin, that puts a vision-language model to work directly on your screen. It looks at what's displayed, works out what you've asked for, and performs the screen interactions the task requires, from taps and swipes to typed input, all on the phone itself, with no PC or USB cable in the loop. The project is essentially a reimplementation of the Python-based mobile agent approach, but built to run on Android and using Shizuku to get system-level input and screenshot permissions without rooting the device. That makes it a practical option for anyone who wants a self-hosted, custom-model alternative to the closed commercial phone assistants. Developers get a multi-agent setup with a tools and skills architecture to build on, while regular users get an app that does the work on a handset they already own.