Repolode
Submit
droidrun

droidrun/mobilerun

An open-source Python framework where LLM agents control Android and iOS phones via natural language, CLI, or Python API.

View on GitHubWebsite
Stars
9,708
Forks
1,034
Open issues
28
Last commit
4 days ago

What is it?

Mobilerun is a framework for letting large language model agents operate mobile devices. Given a natural-language instruction such as opening settings and enabling dark mode, the agent reads the device's UI state, looks at screenshots when needed, and performs taps, swipes, and text input step by step until the task is done.

The framework works with a range of LLM providers, including OpenAI, Anthropic, Gemini, xAI, Ollama, DeepSeek, and OpenRouter, plus OpenAI-compatible endpoints. It offers a CLI for one-off tasks and device checks, Docker for repeatable environments, and a Python API for custom workflows with custom tools and structured output. Android is controlled through a companion Portal app over ADB, and iOS is supported through a separate setup flow.

Optional modes add screenshot-only control for apps without accessibility data, a reasoning mode that separates planning from execution for longer tasks, and tracing through Arize Phoenix or Langfuse.

Key features

  • 01Controls Android and iOS devices with natural language commands
  • 02Works with OpenAI, Anthropic, Gemini, Ollama, DeepSeek and other OpenAI-compatible models
  • 03Combines accessibility trees with screenshots, or uses screenshot-only mode
  • 04Offers a CLI, Docker image and Python API with custom tools
  • 05Supports a reasoning mode for multi-step planning and execution tracing via Phoenix or Langfuse

Use cases

  • Mobile app QA and regression testing driven by plain-language test steps
  • Automating repetitive tasks on one or several phones
  • Extracting data from native mobile apps that lack an API
  • Running mobile workflows from scripts or schedules via the Python API

Quick start

  1. 1

    Install the CLI with uv

    uv tool install mobilerun
  2. 2

    Install the Portal app and enable its accessibility service on your Android device (requires ADB and USB debugging)

    mobilerun setup
  3. 3

    Configure your LLM provider

    mobilerun configure
  4. 4

    Run your first natural language task

    mobilerun run "Open settings and turn on dark mode"

Things to know

  • Requires a paid API key from an LLM provider, or a local model setup.
  • Python 3.14 is not supported; use Python 3.11 to 3.13.
  • Android control needs ADB and a device with USB debugging enabled.

Topics

#ai-agents#llm#mobile-automation#android#ios#python#adb#cli#android-automation#hacktoberfest

FAQ

What is mobilerun?
An open-source Python framework where LLM agents control Android and iOS phones via natural language, CLI, or Python API. Mobilerun is a framework for letting large language model agents operate mobile devices. Given a natural-language instruction such as opening settings and enabling dark mode, the agent reads the device's UI state, looks at screenshots when needed, and performs taps, swipes, and text input step by step until the task is done.
Is mobilerun free to use? What license does it have?
mobilerun is open source under the MIT license. Check the license terms before using it in commercial products.
Is mobilerun still maintained?
Actively maintained. The last commit to mobilerun was on Oct 16, 2026.
What can I use mobilerun for?
Mobile app QA and regression testing driven by plain-language test steps Automating repetitive tasks on one or several phones Extracting data from native mobile apps that lack an API Running mobile workflows from scripts or schedules via the Python API
How do I install mobilerun?
Install the CLI with uv: uv tool install mobilerun Install the Portal app and enable its accessibility service on your Android device (requires ADB and USB debugging): mobilerun setup Configure your LLM provider: mobilerun configure Run your first natural language task: mobilerun run "Open settings and turn on dark mode"
Where can I get mobilerun?
The source code and installation instructions are on GitHub: https://github.com/droidrun/mobilerun

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects