mobilerun

Why we included this project

Mobilerun connects an LLM agent to a phone's accessibility tree and screenshots, so it can inspect the UI, tap, swipe, type, and work through multi-step tasks on real Android and iOS devices, then hand results back through a CLI or Python API. The LLM behind it is not locked in: you can run OpenAI, Anthropic, Gemini, or a local Ollama install, picking whatever fits your cost and privacy needs, and you can drive the whole thing from Docker or the Python API. Teams writing automated QA, onboarding, or device-management scripts are the clearest fit; the toolkit includes a reasoning mode for complex tasks, vision-only control for apps that expose no accessibility data, and tracing hooks for Langfuse or Arize Phoenix. It is a solid choice if you want a controllable, self-hosted agent working on physical devices instead of a simulated environment.