droidrun/mobilerun — reverse-engineered prompt

Build me a Python tool that lets someone control a phone by typing normal language commands, like “open settings and turn on dark mode” or “tell me what app is open.” It should work from the terminal, guide the user through setup, check that a phone is connected, configure an AI provider, then run tasks on Android, with room for iOS support too.

The agent should be able to understand what’s on the screen using the app UI information and screenshots, then tap, swipe, type, launch apps, take screenshots, and return useful results. Please include a simple command line experience, a Python API for people who want to script workflows, and options for harder tasks where the agent can plan several steps before acting.

Keep it provider flexible so people can use common AI models with API keys. Add helpful logs and debugging output so users can see what happened. Look up current docs online if you need to.