Open-source version of 'Doubao Mobile Assistant' arrives: Can operate WeChat in the background without interrupting your phone browsing

Beating AI News Flash: Independent developer Wang Zan has open-sourced an Android AI assistant called Mika, which can directly control apps like WeChat and Xiaohongshu on the phone to help users organize chat histories and send messages.

The author demonstrated having Mika browse WeChat group chats in the background, summarize chat histories, and send messages to WeChat's File Transfer Assistant. Throughout the process, users can continue using other apps on the main screen without having to hand over the screen to the AI.

Doubao Phone Assistant relies on system-level permissions provided by phone manufacturers, whereas Mika launches an auxiliary process with Shell privileges through Android wireless debugging, creating an invisible virtual screen on an ordinary phone where apps like WeChat run. The AI reads buttons, text, and input fields through accessibility services to perform clicks, swipes, and text input; for interfaces where controls cannot be recognized, it uses local OCR or vision models to operate based on visual understanding.

On first use, Mika requires enabling developer options, completing wireless debugging pairing, and authorizing accessibility services. The model's API can be configured independently, and login via a ChatGPT account is also supported. It does not require connecting to a computer during operation, and tools like Codex on a computer can also dispatch tasks to it via MCP.

Cross-app operation functionality is available starting from Android 12. On Android 14, Mika can leverage the virtual screen's independent focus capability to operate in the background without seizing the main screen's input focus.