MobAI

You can harness MobAI's MCP server to automate device control, UI interactions, web automation, and DSL batch execution through any MCP-compatible AI tool. This server provides a robust bridge between your AI assistants and real devices, enabling end-to-end automation with scripts, agents, and web actions.

How to use

Use the MCP server with an MCP client by configuring the client to connect to the local MobAI MCP endpoint. You can list devices, control apps, capture screenshots, execute DSL scripts, and run autonomous AI agents. Web automation lets you interact with mobile browsers and WebViews via CSS selectors, while the DSL executor handles batch automation steps. An AI agent can autonomously complete tasks like opening settings, navigating apps, or performing repetitive actions.

How to install

Prerequisites you need before installing the MCP server include Node.js 18 or newer and the MobAI desktop app running locally to expose the HTTP API on port 8686. You also need a connected Android or iOS device, or an emulator/simulator.

{
  "mcpServers": {
    "mobai": {
      "command": "npx",
      "args": ["-y", "mobai-mcp"]
    }
  }
}

Configuration for MCP clients

To connect common MCP clients such as Cursor, Claude Desktop, Windsurf, or Cline, add the following MCP configuration to each client so it can reach the MobAI MCP server.

{
  "mcpServers": {
    "mobai": {
      "command": "npx",
      "args": ["-y", "mobai-mcp"]
    }
  }
}

Step-by-step setup with common clients

Cursor, Claude Desktop, Windsurf, and other MCP clients can all use the same MCP endpoint configuration. Copy the snippet above into the client’s MCP settings to enable communication with the MobAI MCP server.

Usage patterns with the MCP server

• Start by listing connected devices, then select a device to control and observe its UI structure. - Use UI automation tools to tap, type, and swipe within native apps. - Navigate web content through Web automation to interact with WebViews and browser tabs. - Execute DSL scripts to perform multi-step automation sequences efficiently. - Run autonomous AI agents to tackle complex tasks without manual guidance. - Capture screenshots for verification or reporting.

Security and access considerations

Run the MobAI MCP server locally and restrict access to trusted tools. Use the MobAI desktop app as the trusted gateway, and avoid exposing the MCP API on untrusted networks. If you share access, ensure clients have appropriate permissions for the actions they perform.

Troubleshooting

If you encounter connection issues, verify the MobAI desktop app is running and that the API is accessible at the local address used by your client. If the bridge is not started, run the start_bridge action before attempting automation. On iOS, the bridge may require up to a minute to initialize.

Available tools

list_devices

List all connected devices.

get_device

Get information about a specific device.

start_bridge

Start the on-device automation bridge.

stop_bridge

Stop the automation bridge.

get_screenshot

Capture a screenshot from the device.

get_ui_tree

Retrieve the accessibility/UI tree of the active screen.

tap

Tap a target by index or screen coordinates.

type_text

Type text into an input field.

swipe

Perform a swipe gesture on the screen.

go_home

Return to the device home screen.

launch_app

Launch an app by its bundle/app ID.

list_apps

List installed applications on the device.

execute_dsl

Run a DSL batch automation script.

run_agent

Run an autonomous AI agent to complete complex tasks.

web_list_pages

List open browser tabs and WebViews.

web_navigate

Navigate to a specified URL in a browser/WebView.

web_get_dom

Obtain the DOM tree of a web page.

web_click

Click an element using a CSS selector.

web_type

Type into an element using a CSS selector.

web_execute_js

Execute JavaScript in the current page.

http_request

Make a raw HTTP request to MobAI API.