Autoroid

AI-Powered Android Automation Assistant

👉 Introduction

Autoroid leverages Large Language Models (LLMs) to understand your intent and interact with your Android device through ADB. From navigating apps to automating human actions like taps, swipes, and text input — just tell Autoroid what you want to do.

✨ Features

• Natural Language Control - Ask to do something in human language: Open Settings and enable dark mode

• UI Awareness - Parses Android UI hierarchies to intelligently locate buttons, text, and inputs.

• Universal Android Support - Works with physical devices and emulators as long as ADB is enabled.

• Configurable LLM Backend - Supports OpenAI, Gemini, Anthropic and Ollama.

🚀 Quick Start

1. Install ADB

ADB must be installed and accessible from your terminal.

1. Download Android SDK Platform Tools https://developer.android.com/studio/releases/platform-tools

2. Extract the archive in a new folder adb.

   Example (Windows):

   C:\adb\adb.exe
   

3. Add ADB to PATH

   • Add C:\adb to your system environment variables

4. Connect your device (any one of the below)

   • Enable USB Debugging on your Android device and connect via USB
   • Start an android emulator

5. Verify installation

   adb devices

   You should see your device ID listed.

---

2. Clone the repo

Clone the repository and install dependencies.

git clone https://github.com/qubydev/autoroid.git
cd autoroid

3. Install packages

pip install -r requirements.txt

---

4. Configuration

Create a .env file in the project root.

cp .env.example .env

All Variables

Variable	Description	Example
LLM_PROVIDER	LLM backend provider. Must be compatible with LangChain.	openai | gemini | anthropic
BASE_URL	Custom API base URL. (Leave empty for default)	http://localhost:11434/v1
API_KEY	API key for the selected LLM provider.	sk-...
MODEL	Model name used by the selected LLM provider.	gpt-4o
ADB_PATH	Path to the adb executable if it is not available in your system PATH.	adb
SCREEN_DUMP_PATH	Path on the Android device where the UI hierarchy XML is dumped.	/sdcard/window_dump.xml
LOCAL_DUMP_PATH	Local path where the pulled UI dump XML is stored for processing.	window_dump.xml
RECURSION_LIMIT	Maximum number of reasoning/action steps per goal to prevent infinite loops.	50
RATE_LIMIT_DELAY	Delay in seconds between LLM requests to avoid rate-limit issues.	0.5

---

5. Running Autoroid

Make sure your Android device or emulator is connected.

python main.py

You should see:

🤖 Android Agent Started
Enter Goal:

Example commands:

• Open Chrome and search for cute cats
• Launch Instagram and scroll for 10 seconds
• Enable airplane mode

🤝 Contributing

Contributions are welcome ❤️! Please fork the repository and submit a pull request with your changes.

📄 License

This project is licensed under the MIT License. See the LICENSE file for details.