android-mcp-server
A powerful MCP server that provides comprehensive Android device automation capabilities through ADB, enabling AI agents to interact with Android devices for testing, automation, and device control tasks.
README
Android MCP Server
A powerful MCP (Model Context Protocol) server that provides comprehensive Android device automation capabilities through ADB (Android Debug Bridge). This server enables AI agents and automation tools to interact with Android devices for testing, automation, and device control tasks.
https://github.com/user-attachments/assets/68f3df22-fcf1-49af-8715-fcfa05496c90
🚀 Features
Device Management
- Multi-device support: Connect and manage multiple Android devices simultaneously
- Auto-discovery: Automatic device detection and selection
- Health monitoring: Real-time device status and connectivity checks
- Device information: Complete device specifications and capabilities
UI Automation
- Layout inspection: Extract complete UI hierarchy with element details
- Smart element finding: Search UI elements by text, resource ID, content description, or class
- Gesture control: Tap, swipe, and multi-touch gesture support
- Text input: Send text and key events (back, home, menu, etc.)
Media Capture
- Screenshots: High-quality screen capture with customizable options
- Screen recording: MP4 video recording with quality controls
- Batch operations: Multiple concurrent recording sessions
Advanced Monitoring
- Live log monitoring: Real-time logcat streaming with filtering
- Multi-session support: Concurrent log monitoring with different filters
- Export capabilities: Save logs and media to local filesystem
📋 Prerequisites
Required Software
- Python 3.11+: Runtime environment
- Android SDK Platform Tools: For ADB functionality
- USB Debugging enabled: On target Android devices
ADB Setup
-
Install Android SDK Platform Tools:
# macOS (via Homebrew) brew install android-platform-tools # Ubuntu/Debian sudo apt install android-tools-adb # Windows # Download from: https://developer.android.com/tools/releases/platform-tools -
Verify ADB installation:
adb version # Should output: Android Debug Bridge version x.x.x -
Enable USB Debugging on Android device:
- Go to Settings → About Phone
- Tap "Build number" 7 times to enable Developer Options
- Go to Settings → Developer Options
- Enable "USB Debugging"
- Connect device via USB and authorize when prompted
⚙️ Installation
Using uv (Recommended)
git clone <repository-url>
cd android-mcp-server
uv sync
Using pip
git clone <repository-url>
cd android-mcp-server
pip install -e .
Verify Installation
# Check if devices are detected
adb devices
# Test server startup
python -m src.server
🚀 Quick Start
Basic Usage Example
import asyncio
from mcp import ClientSession
async def demo_automation():
# Connect to MCP server
session = ClientSession()
# Get connected devices
devices = await session.call_tool("get_devices")
print(f"Found {devices['count']} devices")
# Take a screenshot
screenshot = await session.call_tool("take_screenshot", {
"filename": "demo_screenshot.png"
})
# Find and tap a button
elements = await session.call_tool("find_elements", {
"text": "Settings",
"clickable_only": True
})
if elements["count"] > 0:
result = await session.call_tool("tap_element", {
"text": "Settings"
})
print("Tapped Settings button")
# Run the demo
asyncio.run(demo_automation())
Command Line Usage
# Start the MCP server
python -m src.server
# Or using the installed script
android-mcp-server
🛠️ Available Tools
Device Management
get_devices(): List all connected Android devicesselect_device(device_id?): Select specific device or auto-select first availableget_device_info(): Get detailed device information and screen specifications
UI Interaction
get_ui_layout(compressed?, include_invisible?): Extract complete UI hierarchyfind_elements(...): Find UI elements by text, resource ID, content description, or classtap_screen(x, y): Tap at specific screen coordinatestap_element(text?, resource_id?, content_desc?, index?): Find and tap UI elementswipe_screen(start_x, start_y, end_x, end_y, duration_ms?): Swipe between coordinatesswipe_direction(direction, distance?, duration_ms?): Swipe in direction (up/down/left/right)input_text(text, clear_existing?): Input text into focused fieldpress_key(keycode): Press device keys (BACK, HOME, ENTER, etc.)
Media Capture
take_screenshot(filename?, pull_to_local?): Capture device screenshotstart_screen_recording(...): Start screen recording with quality optionsstop_screen_recording(recording_id?, pull_to_local?): Stop recording sessionlist_active_recordings(): List all active recording sessions
Log Monitoring
get_logcat(tag_filter?, priority?, max_lines?, clear_first?): Get device logsstart_log_monitoring(tag_filter?, priority?, output_file?): Start continuous log monitoringstop_log_monitoring(monitor_id?): Stop log monitoring sessionlist_active_monitors(): List active monitoring sessions
🔌 MCP Server Installation
Add to Claude Desktop (Recommended)
-
Locate your Claude Desktop configuration file:
# macOS ~/Library/Application Support/Claude/claude_desktop_config.json # Windows %APPDATA%/Claude/claude_desktop_config.json -
Add the Android MCP Server:
{ "mcpServers": { "android-mcp-server": { "command": "uv", "args": [ "--directory", "/path/to/android-mcp-server", "run", "android-mcp-server" ] } } } -
Restart Claude Desktop to activate the server.
Alternative Installation Methods
Using Python directly
{
"mcpServers": {
"android-mcp-server": {
"command": "python",
"args": ["-m", "src.server"],
"cwd": "/path/to/android-mcp-server"
}
}
}
Using pip installed package
pip install -e /path/to/android-mcp-server
{
"mcpServers": {
"android-mcp-server": {
"command": "android-mcp-server"
}
}
}
🤖 Usage Examples for AI Agents
Basic Device Interaction
"Take a screenshot of the current Android device screen and then find all buttons containing the word 'Settings'"
"Connect to the Android device, open the Settings app by tapping on it, then take another screenshot"
"Swipe down from the top of the screen to open the notification panel, then capture what's shown"
App Testing and Automation
"Launch the Calculator app, perform the calculation 25 + 37, and verify the result is 62"
"Test the login flow: find the username field, enter 'testuser', find the password field, enter 'password123', then tap the login button"
"Navigate through the app's main menu by tapping each menu item and taking screenshots of each page"
UI Analysis and Debugging
"Extract the complete UI layout of the current screen and identify all clickable elements"
"Find all text input fields on this screen and tell me which ones are currently focused"
"Look for any accessibility issues by checking if buttons have proper content descriptions"
Log Monitoring and Debugging
"Start monitoring logs for the 'MyApp' package and show me any errors that occur in the next 2 minutes"
"Get the last 50 lines of system logs with priority level Warning or higher"
"Monitor network-related logs while I perform the next action, then show me what was logged"
Performance Testing
"Record a 30-second video of the app launch sequence, then analyze the UI responsiveness"
"Take screenshots before and after each major user action to document the user flow"
"Monitor memory usage logs while navigating through different screens of the app"
Multi-Device Testing
"List all connected Android devices and select the one with the largest screen resolution"
"Switch to device ID 'emulator-5554' and repeat the previous test sequence"
"Compare the same UI element across different devices by taking screenshots on each"
💡 Tips for AI Agents
Efficient Element Finding
- Always use
find_elementsbeforetap_elementto verify elements exist - Use
exact_match: falsefor partial text matching when element text might vary - Combine multiple search criteria (text + clickable_only) for more precise targeting
Context-Aware Log Monitoring
- Limit
max_linesto 20-200 to avoid overwhelming the AI context window - Use specific
tag_filterto focus on relevant app components - Set appropriate
prioritylevel (I/W/E) based on the type of information needed
Screenshot-Driven Workflows
- Take screenshots before and after actions for verification
- Use descriptive filenames that indicate the test step or app state
- Screenshots are automatically saved to
./assets/for easy access
Robust Automation Patterns
- Check device connection with
get_device_infobefore starting automation - Use
get_ui_layoutwhen element finding fails to understand the current screen structure - Implement retry logic for intermittent UI state issues
🏗️ Architecture
The server is built with a modular architecture:
ADBManager: Core ADB communication and device managementUILayoutExtractor: UI hierarchy parsing and element extractionScreenInteractor: Touch and gesture handlingMediaCapture: Screenshot and recording functionalityLogMonitor: Real-time log monitoring and filteringFastMCP: MCP protocol implementation and tool registration
🤝 Contributing
Contributions are welcome! Please feel free to submit issues and pull requests.
Development Setup
git clone <repository-url>
cd android-mcp-server
uv sync --dev
Running Tests
# Install test dependencies
pip install -e ".[test,dev]"
# Run full test suite with coverage
python -m pytest tests/ --cov=src --cov-fail-under=80
# Run specific test categories
python -m pytest tests/ -m "unit" # Unit tests only
python -m pytest tests/ -m "integration" # Integration tests
python -m pytest tests/ -m "not slow" # Skip slow tests
# Run code quality checks
make code-quality # or run individual tools:
flake8 src
pydocstyle src --convention=google
📄 License
This project is licensed under the MIT License - see the LICENSE file for details.
🔗 Related Projects
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。