•
Drizz raises $2.7M in seed funding •
•
Featured on Forbes
•
Drizz raises $2.7M in seed funding •
•
Featured on Forbes

Appium MCP is a Model Context Protocol server that lets an AI assistant such as Cursor, Claude Code or Gemini CLI drive Android and iOS devices through Appium. You describe what you want in the chat, and the assistant starts the session, inspects the screen and performs the steps by calling Appium for you.
This guide covers what Appium MCP is, which of the several projects with that name people usually mean, how to set it up, what it is good for, and where it stops. If you need the basics of Appium itself first, start with what Appium is and how its architecture works.
The Model Context Protocol (MCP) is an open standard that lets AI assistants call external tools. An MCP server exposes a set of tools, and the assistant decides when to call them while it works on your request.
An Appium MCP server exposes Appium automation as those tools. The assistant can start a session on a device, find elements, tap and type, and generate locators and test code from what it sees.
The official implementation is MCP Appium, maintained under the Appium project at appium/appium-mcp and published on npm as appium-mcp. It describes itself as an MCP server for mobile development and automation across iOS, Android, simulators, emulators and real devices.
Several projects share the name, which makes search results confusing. These are the ones you are most likely to run into.
| Project | Maintained by | Install | What to know |
|---|---|---|---|
MCP Appium (appium-mcp on npm) |
Appium project | npx -y appium-mcp@latest |
The official server. The rest of this guide covers this one. |
mcp-appium-visual |
Community (Rahul Sharma) | npx mcp-appium-visual |
Adds OCR, image matching and visual element detection on top of Appium. Needs Appium 2 installed; the default setup expects you to start it yourself. |
appium-mcp on PyPI |
Community (You Can Automate) | pip install appium-mcp |
A separate Python project with YAML workflows and AWS Bedrock integration. Needs a running Appium server. |
appium-mcp-server |
Community (argneshu) | npx -y appium-mcp-server@latest |
Needs a running Appium server on port 4723 and Python 3.12 or later. |
Note that appium-mcp on npm and appium-mcp on PyPI are different projects from different maintainers.
ANDROID_HOME setCursor: go to Cursor Settings, then MCP, then Add new MCP Server, and use the command npx -y appium-mcp@latest.
Claude Code:
claude mcp add appium-mcp -- npx -y appium-mcp@latestGemini CLI:
gemini mcp add appium-mcp npx -y appium-mcp@latestAny other MCP client that reads a JSON config:
{
"mcpServers": {
"appium-mcp": {
"command": "npx",
"args": ["appium-mcp@latest"],
"env": {
"ANDROID_HOME": "/path/to/Android/sdk"
}
}
}
}MCP Appium supports two driver modes.
remoteServerUrl when creating or attaching to a session and it connects to an Appium server you already run instead. This mode is also required for platforms other than Android and iOS.With an emulator or simulator running, ask the assistant something like:
Start an Android session on the running emulator, open the app, and tap Login. Then list the elements on the next screen.
The project is moving quickly, so check the README in the repository for the current tool list and options before you rely on any one of them.
These are two different uses of AI.
Appium MCP puts an AI assistant in front of Appium. The assistant writes and operates selector-based automation faster than a person would, but the result is still selector-based automation.
AI-native tools such as Drizz replace the selector layer. Tests are written in plain English and Vision AI finds each element on the rendered screen, so there is no locator to maintain when the UI changes.
Which one helps depends on where your time goes. If the bottleneck is writing Appium tests, Appium MCP will speed that up. If the bottleneck is maintaining them, it will not change much, and it is worth comparing the best Appium alternatives or reading Drizz vs Appium.
It is a Model Context Protocol server that exposes Appium's mobile automation as tools an AI assistant can call. It lets the assistant start device sessions, find and interact with elements, and generate locators and test code.
Yes. MCP Appium, published on npm as appium-mcp, is maintained under the Appium project. Several community projects use similar names, so check the maintainer and package registry before installing.
Not for local sessions with the official server. It bundles the UiAutomator2 and XCUITest drivers and uses them directly. Some community servers do require you to start Appium yourself.
No. It can generate locators for you, but the tests still depend on them and still break when the underlying identifiers change.
Any MCP-compatible client. The project documents setup for Cursor, Claude Code and Gemini CLI.