MCP Desktop Automation

Desktop Automation MCP Server lets you control the mouse, keyboard, and screen captures from an MCP client. It enables large language models to perform desktop tasks such as moving the cursor, clicking, typing text, and taking screenshots, making it possible to automate GUI workflows and testing.

How to use

You use the Desktop Automation MCP Server by connecting your MCP client to the server via the supported runtime method. The server exposes a set of simple actions for GUI automation, including moving the mouse, clicking, typing, and capturing screenshots. You can build automated workflows that perform complex desktop interactions by sequencing these actions.

Key actions you can perform include moving the mouse to specific coordinates, clicking (left, right, or middle), pressing keys or key combinations, typing text at the current cursor, and capturing screenshots. Screenshots can be listed and retrieved by their IDs for later use in validation or reporting.

To configure the server for use from an MCP client, you typically run it via a local runtime such as NPX, which starts the server process from a package name. The following configuration shows how to connect using NPX and the canonical package name for the Desktop Automation server.

{
  "mcpServers": {
    "desktop_automation": {
      "command": "npx",
      "args": ["-y", "mcp-desktop-automation"]
    }
  }
}

Available tools

get_screen_size

Returns the current screen width and height with no inputs.

screen_capture

Captures the current desktop screen and returns image data.

keyboard_press

Simulates pressing a keyboard key or key combination. Requires a key and optional modifiers.

keyboard_type

Types a string of text at the current cursor position.

mouse_click

Performs a mouse click with optional button and double-click mode.

mouse_move

Moves the mouse to specified X and Y coordinates.

screenshot_list

Lists all available screenshots by name.

screenshot_content

Retrieves a PNG image for a given screenshot ID.