appium-mcp-server
Provides Appium mobile automation capabilities to Claude and other MCP-compatible clients, enabling cross-platform iOS and Android automation, element interaction, and session management.
README
Appium MCP Server
A Model Context Protocol (MCP) server that provides Appium mobile automation capabilities to Claude and other MCP-compatible clients.
Features
- Cross-platform mobile automation: Support for both iOS and Android
- Multiple app launch modes: Native apps, browser automation, or installed apps
- Element interaction: Find, tap, and input text to mobile elements
- Session management: Start, monitor, and terminate Appium sessions
- Page inspection: Get page source and scroll functionality
- Easy integration: Works seamlessly with Claude Desktop and other MCP clients
Prerequisites
- Python 3.12+ installed on your system : Environment variable should be set correctly in your bash or zsh so that claude can read it from your system
- Node.js 18.1+ (for npx usage)
- Appium Server running on
http://localhost:4723 - Mobile devices/simulators configured and accessible
Setting up Appium Server
# Install Appium globally
npm install -g appium
# Install drivers for your platforms
appium driver install xcuitest # for iOS
appium driver install uiautomator2 # for Android
# Start Appium server
appium server --port 4723
Installation & Usage
Option 1: Using npx
Add this configuration to your Claude Desktop config file:
{
"mcpServers": {
"appium-mcp-server": {
"command": "npx",
"args": ["-y", "appium-mcp-server@latest"]
}
}
}
Requirements Node.js >= 18 installed globally npx available in PATH
🔹 2. ⚠️ Important note (THIS IS KEY)
> ⚠️ Note:
> If you are using Node via NVM, Claude Desktop may not detect it automatically.
> In that case, use Option 2 below.
### Option 2: NVM Users (Recommended if Option 1 fails)
```json
{
"mcpServers": {
"appium-mcp-server": {
"command": "bash",
"args": [
"-lc",
"export NVM_DIR=\"$HOME/.nvm\" && source \"$NVM_DIR/nvm.sh\" && npx -y appium-mcp-server@latest"
]
}
}
}
Option 3: NVM Users (Recommended if Option 2 fails)
{
"mcpServers": {
"appium-mcp-server": {
"command": "bash",
"args": [
"-lc",
"export NVM_DIR=\"$HOME/.nvm\" && [ -s \"/opt/homebrew/opt/nvm/nvm.sh\" ] && . \"/opt/homebrew/opt/nvm/nvm.sh\" && nvm use 20 >/dev/null && npx -y appium-mcp-server@latest"
]
}
}
}
🔹 4. 💪 Advanced / stable (BEST for reliability)
### Option 3: Use Node directly (Most reliable)
```json
{
"mcpServers": {
"appium-mcp-server": {
"command": "/path/to/your/node",
"args": [
"/path/to/appium-mcp-server/bin/appium-mcp-server.js"
]
}
}
}
To find your Node path: which node
🧠 Why this is the BEST approach
You cover:
| User type | Works? |
|---|---|
| Beginner | ✅ npx |
| NVM user | ✅ bash fix |
| Power user | ✅ direct node |
Option 4: VS Code Local Setup (Github Copilot / Cursor)
If you want to use appium-mcp-server with GitHub Copilot or Cursor inside your existing VS Code project, follow these steps:
At the root of your existing project, create a .vscode folder (if it doesn't exist) and inside it create a file named mcp.json:
your-existing-project/
└── .vscode/
└── mcp.json
Paste the following into mcp.json:
{
"servers": {
"appium-mcp-prod": {
"command": "/Users/username/appium-mcp-server/.venv/bin/python",
"args": ["src/mcp_server.py"],
"cwd": "~/appium-mcp-server",
"env": {
"APPIUM_PORT": "4723"
}
}
}
}
⚠️ Replace
/Users/username/with the actual path where you clonedappium-mcp-server.
Runpwdinside the cloned folder to get the exact path.
⚙️ Local System Setup Instructions for Apple M1/M2 and Windows
🍎 Apple Silicon (M1/M2) – macOS
Apple Silicon users must rebuild the Python virtual environment (.venv) locally to avoid architecture compatibility errors like:
ImportError: ... incompatible architecture (have 'x86_64', need 'arm64')
✅ Steps for M1/M2 Macs:
- 🧬 Clone the project
- 📂 Open Terminal
- Navigate to your project repository
- 🛠️ Run:
chmod +x bootstrap.sh - 🛠️ Run:
./bootstrap.sh- 🧹 Removes the prebuilt
.venv - 🧱 Recreates
.venvusing nativearm64Python - 📦 Reinstalls all Python dependencies
- 🧹 Removes the prebuilt
- 🔁 You only need to do this once, unless
.venvis deleted orrequirements.txtchanges
💻 Windows Users
Windows users can use the bundled .venv if compatible, or regenerate it locally.
✅ Steps for Windows:
- 🧬 Clone the project
- 📂 Open Command Prompt or PowerShell
- 📁 Navigate to your project folder (e.g.
cd C:\Users\YourName\appium-mcp-server) - 🛠️ Run:
bootstrap.bat- 🧱 Creates a fresh
.venvusing your system’s Python (≥ 3.10) - 📦 Installs all dependencies from
requirements.txt
- 🧱 Creates a fresh
- 🛑 Make sure Python is in your
PATHand is version ≥ 3.10
🧪 Verify It Works
🧪 Starting the Server
After running the appropriate setup script (./bootstrap.sh on macOS or bootstrap.bat on Windows), start the MCP server using either of the following:
npx appium-mcp-server
Or run it directly from project path:
node bin/appium-mcp-server.js
You should see output like:
🚀 Starting MCP server using python3.12
🔧 Injecting PYTHONPATH = ...
### 🧪 Using Claude Desktop with a Local Project
To run your **local version** of the MCP server with Claude Desktop, follow these steps:
1. Open **Claude Desktop**
2. Go to **Settings → Developer → EditConfig button
3. This will open the `desktip-claude-config.json` file
4. Add the following configuration under `"mcpServers"`:
---
#### 🍎 macOS (Intel or Apple Silicon)
```json
{
"mcpServers": {
"local-appium-mcp": {
"command": "node",
"args": ["/Users/your.name/appium-mcp-server/bin/appium-mcp-server.js"]
}
}
}
💻 Windows
{
"mcpServers": {
"local-appium-mcp": {
"command": "node",
"args": ["C:\\Users\\your.name\\appium-mcp-server\\bin\\appium-mcp-server.js"]
}
}
}
📝 Replace /Users/your.name/... with the full path to your cloned project directory.
## Available Tools
### Session Management
- **`appium_start_session`**: Start an Appium session
- `platform`: "iOS" or "Android"
- `device_name`: Device name or UDID
- `app_path`: Path to app file (optional)
- `bundle_id`: iOS bundle ID (optional)
- `app_package`: Android package name (optional)
- `app_activity`: Android activity (optional)
- **`appium_get_session_info`**: Get current session information
- **`appium_quit_session`**: Terminate the current session
### Element Interaction
- **`appium_find_element`**: Find elements on screen
- `strategy`: "id", "xpath", "class_name", or "accessibility_id"
- `value`: Locator value
- **`appium_tap_element`**: Tap on an element
- `element_id`: ID of previously found element
- **`appium_input_text`**: Send text to input fields
- `element_id`: Target element ID
- `text`: Text to input
### Page Navigation
- **`appium_get_page_source`**: Get current page source
- **`appium_scroll`**: Scroll the screen
- `direction`: "up" or "down"
## Example Usage with Claude
Start an iOS session on iPhone 15 Pro Max simulator, then navigate to the SauceDemo website and automate the login process.
Claude will use the MCP server to:
1. Start an Appium session
2. Find login elements
3. Input credentials
4. Navigate through the app
## Configuration Examples
### iOS Safari Browser Testing
```json
{
"platform": "iOS",
"device_name": "iPhone 15 Pro Max"
}
Android Chrome Browser Testing
{
"platform": "Android",
"device_name": "Android Emulator"
}
iOS Native App Testing
{
"platform": "iOS",
"device_name": "iPhone 15 Pro Max",
"bundle_id": "com.example.myapp"
}
Android Native App Testing
{
"platform": "Android",
"device_name": "Android Emulator",
"app_package": "com.example.myapp",
"app_activity": ".MainActivity"
}
🚀 Quick Start with Gemini CLI
The easiest way to run mobile automation commands is using our simple CLI wrapper with Gemini.
Setup Instructions
- 🧬 Clone the project
- 📂 Open Terminal
- Navigate to your project repository
- Also make sure you have created .env file in your project root and entered GEMINI_API_KEY key there
Load environment variables from .env
load_dotenv()
Get the API key
api_key = os.getenv("GEMINI_API_KEY")
-
Make the setup script executable:
chmod +x mobile-setup.sh -
Load the mobile function:
source mobile-setup.shYou should see:
Mobile automation function loaded! Usage: mobile "Launch Settings on iPhone" mobile -i mobile --claude "Open Instagram" To make this permanent, add this to your ~/.bashrc or ~/.zshrc: source "/path/to/your/mobile-setup.sh" -
Test the function:
# Test help mobile -h # Test interactive mode mobile -i # Test single command mobile "Launch Settings on iPhone" # Test with Claude mobile --claude "Open Instagram and scroll down"
Make it Permanent (Optional)
To have the mobile function available every time you open a terminal:
For Bash users:
echo "source $(pwd)/mobile-setup.sh" >> ~/.bashrc
For Zsh users:
echo "source $(pwd)/mobile-setup.sh" >> ~/.zshrc
For Fish shell users:
mkdir -p ~/.config/fish/functions
# Then manually create the fish function file
Usage Examples
# Single prompts (default Gemini)
mobile "Launch Settings on iPhone 15 Pro Max"
mobile "Open Instagram and like the first post"
mobile "Calculate 15 + 25 in Calculator app"
mobile "Launch Safari and go to google.com"
# Single prompts with Claude
mobile --claude "Open Notes and create a new note"
mobile --claude "Launch Camera and take a photo"
# Interactive modes
mobile -i # Interactive with Gemini
mobile --claude -i # Interactive with Claude
mobile --interactive # Interactive with Gemini
# Help
mobile -h
mobile --help
For Windows:
Run the setup script: cmdmobile-setup.bat You should see: Mobile automation command created!
Usage: mobile "Launch Settings on iPhone" mobile -i mobile --claude "Open Instagram"
The 'mobile' command is now available for this session. You should see: Mobile automation function loaded! Usage: mobile "Launch Settings on iPhone" mobile -i mobile --claude "Open Instagram"
To make this permanent, add this to your ~/.bashrc or ~/.zshrc: source "/path/to/your/mobile-setup.sh"
Test the function: bash# Test help mobile -h
Test interactive mode
mobile -i
Test single command
mobile "Launch Settings on iPhone"
Test with Claude
mobile --claude "Open Instagram and scroll down"
Make it Permanent (Optional) For Windows: Command Prompt:
Add the script directory to your system PATH, or Run mobile-setup.bat each time you open a new command prompt
Usage Examples bash# Single prompts (default Gemini) mobile "Launch Settings on iPhone 15 Pro Max" mobile "Open Instagram and like the first post" mobile "Calculate 15 + 25 in Calculator app" mobile "Launch Safari and go to google.com"
Single prompts with Claude
mobile --claude "Open Notes and create a new note" mobile --claude "Launch Camera and take a photo"
Interactive modes
mobile -i # Interactive with Gemini mobile --claude -i # Interactive with Claude mobile --interactive # Interactive with Gemini
Help
mobile -h mobile --help
Benefits
✅ No PATH modifications needed
✅ Self-contained in your project directory
✅ Works from any directory once loaded
✅ Easy to modify and customize
✅ No external files created
✅ Portable - just copy the script
Quick Workflow
-
One-time setup:
chmod +x mobile-setup.sh source mobile-setup.sh -
Daily usage:
mobile "Launch Settings" mobile "Open Instagram" mobile -i # for interactive mode -
For permanent access (optional):
echo "source $(pwd)/mobile-setup.sh" >> ~/.bashrc # Then restart terminal or run: source ~/.bashrc
Troubleshooting
Common Issues
-
"No active session" errors
- Ensure Appium server is running on localhost:4723
- Check device/simulator availability
- Verify platform-specific setup (Xcode for iOS, Android SDK for Android)
-
Element not found errors
- Use
appium_get_page_sourceto inspect available elements - Try different locator strategies (id, xpath, accessibility_id)
- Ensure proper timing - elements may need time to load
- Use
-
Python environment issues
- The package automatically creates a virtual environment
- If issues persist, check Python 3.8+ installation
- Verify internet connectivity for dependency installation
Debug Mode
Set DEBUG=1 environment variable for verbose logging:
DEBUG=1 npx appium-mcp-server
Development
To modify or contribute to this package:
git clone https://github.com/yourusername/appium-mcp-server.git
cd appium-mcp-server
npm install
npm link # for local testing
License
MIT License - see LICENSE file for details.
Contributing
Contributions welcome! Please read the contributing guidelines and submit pull requests to the main repository.
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。