Android MCP Server
An MCP server designed for Android development, enabling AI assistants to directly control Android devices for screenshots, UI analysis, app management, and more.
README
Android MCP Server
🤖 An MCP (Model Context Protocol) server designed for Android development, enabling AI assistants to directly control Android devices for screenshots, UI analysis, app management, and more.
Table of Contents
Introduction
This project is built on the FastMCP framework and communicates with Android devices via ADB (Android Debug Bridge). It exposes 19 practical tools to AI assistants (such as Claude, CodeBuddy, etc.), covering the complete Android development and debugging workflow:
Screenshot → UI Analysis → Interaction → Logcat → App Management → File Transfer
Prerequisites
- Python >= 3.11
- uv package manager
- ADB installed and added to PATH (
adb --versionworks) - Android device connected or emulator running (
adb devicesrecognizes it) - USB Debugging enabled on the Android device
Installation
# Clone the project
git clone https://github.com/huarangmeng/AndroidMcpServer
cd AndroidMcpServer
# Install dependencies
uv sync
Dependencies
| Package | Version | Purpose |
|---|---|---|
mcp[cli] |
>= 1.2.0 | MCP framework |
Pillow |
>= 10.3.0 | Screenshot processing |
Configuration
CodeBuddy
Add the following to ~/.codebuddy/mcp.json:
{
"mcpServers": {
"android_mcp": {
"command": "/Users/<your-username>/.local/bin/uv",
"args": [
"run",
"--project", "/path/to/AndroidMcpServer",
"python",
"/path/to/AndroidMcpServer/main.py",
"--mode", "stdio",
"--temp-dir", "/tmp/android_mcp"
],
"transportType": "stdio"
}
}
}
Claude Desktop
Add to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"android_mcp": {
"command": "uv",
"args": [
"run",
"--project", "/path/to/AndroidMcpServer",
"python",
"/path/to/AndroidMcpServer/main.py",
"--mode", "stdio",
"--temp-dir", "/tmp/android_mcp"
]
}
}
}
HTTP Mode
# Start HTTP server
uv run main.py --mode streamable-http --temp-dir /tmp/android_mcp --port 3001
# Service URL: http://localhost:3001/mcp
Tools
A total of 19 tools in 5 categories:
📸 Screenshot & UI Analysis
| Tool | Description |
|---|---|
get_screenshot |
Capture device screenshot (auto-scaled to 50%) |
get_ui_dump |
Get XML UI hierarchy with attribute filtering |
🖱️ Interaction
| Tool | Parameters | Description |
|---|---|---|
tap_screen |
x, y |
Tap at coordinates |
swipe_screen |
x1,y1,x2,y2,duration_ms |
Swipe gesture |
long_press |
x, y, duration_ms |
Long press |
send_text |
text_to_send |
Send text to focused input (English only) |
perform_system_action |
action |
System keys: BACK / HOME / RECENT_APPS |
📱 App Management
| Tool | Parameters | Description |
|---|---|---|
launch_app |
package_name |
Launch app |
force_stop_app |
package_name |
Force stop app |
clear_app_data |
package_name |
Clear app data |
install_apk |
apk_path, replace_existing |
Install APK |
get_app_list |
include_system_apps |
List installed apps |
grant_permission |
package_name, permission |
Grant runtime permission |
🔍 Debugging & Diagnostics
| Tool | Parameters | Description |
|---|---|---|
get_logcat_output |
app_package, log_level |
Get last 100 logcat lines |
get_device_info |
— | Get device model, OS version, etc. |
get_current_activity |
— | Get current foreground Activity |
get_memory_info |
package_name |
View app memory usage |
📂 File Operations
| Tool | Parameters | Description |
|---|---|---|
pull_file |
device_path, local_path |
Pull file from device |
push_file |
local_path, device_path |
Push file to device |
Debugging
Use MCP Inspector for local debugging:
# Install MCP Inspector
npm install -g @modelcontextprotocol/inspector
# Start with config file
npx @modelcontextprotocol/inspector --config mcp-inspector-config.json --server android-stdio
Or start manually:
# stdio mode
uv run main.py --mode stdio --temp-dir /tmp/android_mcp
# HTTP mode
uv run main.py --mode streamable-http --temp-dir /tmp/android_mcp --port 3001
References
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。