Android-MCP
服务介绍
Android-MCP
Android-MCP is a lightweight, open-source tool that bridges between AI agents and Android devices. Running as an MCP server, it lets LLM agents perform real-world tasks such as app navigation, UI interaction and automated QA testing without relying on traditional computer-vision pipelines or preprogrammed scripts.
Forked and customized by Hady Ahmed
Features
- Direct Device Control: Click, swipe, drag, type, and press buttons on Android devices
- UI State Inspection: Get device state with UI hierarchy and optional annotated screenshots
- MCP Integration: Works with any MCP-compatible client (Claude Desktop, VS Code, etc.)
- Emulator Support: Works with Android emulators (tested on emulator-5554)
- Physical Device Support: Connect to real Android devices via ADB
- Vision Capabilities: Generate annotated screenshots with numbered UI elements for vision-based AI agents
Requirements
- Python 3.12+
- Android device or emulator running
- ADB (Android Debug Bridge) installed and configured
uiautomator2compatible Android device (Android 4.4+)
Installation
From Source
- Clone the repository:
git clone https://github.com/HadyAhmed00/Android-MCP.git
cd Android-MCP
- Install dependencies:
pip install -e .
Or install manually:
pip install mcp uiautomator2 pillow ipykernel
Quick Start
1. Start the MCP Server
For emulator:
python main.py --emulator
For physical device:
python main.py
The server will connect to your Android device via ADB and start listening for MCP client connections.
2. Configure with MCP Client
To use this with Claude Code or other MCP clients, add the following to your MCP configuration:
Example MCP Config:
{
"mcpServers": {
"android-mcp": {
"command": "python",
"args": [
"/path/to/Android-MCP/main.py",
"--emulator"
]
}
}
}
Available Tools
The MCP server exposes 9 tools for controlling Android devices:
1. State-Tool
Get the current state of the device including UI hierarchy and optional screenshot.
Parameters:
use_vision(bool, optional): Include annotated screenshot with labeled UI elements
Example:
Get device state with screenshot
2. Click-Tool
Click on a specific coordinate on the screen.
Parameters:
x(int): X coordinatey(int): Y coordinate
Example:
Click on coordinates 540, 800
3. Long-Click-Tool
Long press (hold) on a specific coordinate.
Parameters:
x(int): X coordinatey(int): Y coordinate
Example:
Long click on 540, 800 for 2 seconds
4. Swipe-Tool
Perform a swipe gesture from one point to another.
Parameters:
x1(int): Starting X coordinatey1(int): Starting Y coordinatex2(int): Ending X coordinatey2(int): Ending Y coordinate
Example:
Swipe from top to bottom (refresh)
5. Type-Tool
Type text at a specific coordinate (automatically focuses the field).
Parameters:
text(str): Text to typex(int): X coordinatey(int): Y coordinateclear(bool, optional): Clear existing text before typing
Example:
Type "hello world" into the search field
6. Drag-Tool
Drag from one location and drop at another.
Parameters:
x1(int): Starting X coordinatey1(int): Starting Y coordinatex2(int): Ending X coordinatey2(int): Ending Y coordinate
Example:
Drag and drop item from position to trash
7. Press-Tool
Press device buttons (home, back, power, volume, etc.).
Parameters:
button(str): Button name (back, home, power, volume_up, volume_down)
Example:
Press the back button
8. Notification-Tool
Open the notification bar to access notifications.
Parameters: None
Example:
Open notification bar
9. Wait-Tool
Wait for a specified duration (useful for allowing apps to load).
Parameters:
duration(int): Seconds to wait
Example:
Wait for 2 seconds
Usage Workflow
Basic Example: Navigate and Click
-
Get device state with vision:
Get the current state of the device with a screenshotThis returns the UI hierarchy and an annotated screenshot showing numbered UI elements.
-
Click on an element:
Click on element 0 (the button that was labeled as 0) -
Get updated state:
Get the device state again
Example: Form Submission
1. Get device state with vision to see the form
2. Click on the name field (element 2)
3. Type "John Doe"
4. Click on the email field (element 3)
5. Type "john@example.com"
6. Scroll down to find the submit button
7. Click the submit button
8. Wait 2 seconds for the form to process
9. Get device state to confirm submission
Example: App Navigation
1. Get device state to see current screen
2. Click the "Settings" button
3. Wait 1 second for settings to load
4. Get device state to see settings menu
5. Click on "About Phone"
6. Get state to view device information
Architecture
The project is organized into three main modules:
main.py
- Entry point that creates the FastMCP server
- Defines and exposes all 9 tools
- Handles command-line arguments (
--emulator)
src/mobile/
- Mobile class: Manages device connection and state
- MobileState: Data class representing device state
- Captures screenshots and processes them
src/tree/
- Tree class: Parses Android UI XML hierarchy
- Extracts interactive elements (buttons, inputs, etc.)
- Generates annotated screenshots with numbered labels
- Helper utilities for coordinate extraction
Troubleshooting
Device Not Connecting
-
Check ADB:
adb devicesYour device should appear in the list.
-
For emulators:
adb connect emulator-5554 -
Enable USB Debugging (for physical devices):
- Go to Settings > Developer Options
- Enable USB Debugging
- Connect your device
No Interactive Elements Found
This can happen if:
- The app hasn't fully loaded - try waiting with
Wait-Tool - The UI uses unusual widgets not in the
INTERACTIVE_CLASSESlist - Screen is at an unusual rotation
Try getting device state with vision to see what's actually on screen.
Screenshots Not Working
- Ensure the device screen is on
- Check that the device has enough storage
- Try waiting a moment before requesting screenshots
Development
Running Tests
Tests for the UI tree parsing:
pytest tests/
Adding New Tools
To add a new tool to the MCP server:
- Add a function in
main.pydecorated with@mcp.tool() - Define parameters with type hints
- Return results that can be serialized (strings, lists, or Images)
Example:
@mcp.tool(name='My-Tool', description='Description of what this does')
def my_tool(param1: str, param2: int):
result = mobile.get_device().some_action(param1, param2)
return f'Action completed: {result}'
License
This project is licensed under the MIT License - see LICENSE file for details.
Contributing
Contributions are welcome! Please see CONTRIBUTING.md for guidelines.
Support
If you encounter issues:
- Check the Troubleshooting section above
- Verify your Android device setup with ADB
- Open an issue on GitHub with:
- Device details (type, Android version)
- Error messages
- Steps to reproduce