Windows-MCP

Windows-MCP is a lightweight, open-source server that connects AI agents to the Windows operating system. It enables agents to navigate files, control applications, interact with the UI, QA test, and automate a wide range of Windows tasks with real-time responsiveness.

How to use

You use Windows-MCP by connecting your MCP-compatible client to the local MCP server instance. The server exposes a set of tools that let your agent click, type, scroll, drag, move the mouse, press keyboard shortcuts,Wait, capture window and desktop state, launch apps, and run PowerShell commands. Your agent can then orchestrate multi-step workflows across Windows UI elements, automate QA tests, or perform routine maintenance tasks. Start by configuring the client to point at the local server and then issue high-level actions like “open File Explorer,” “type search text,” or “activate a window” to drive your automation flow.

How to install

Prerequisites you need before installing Windows-MCP are Python 3.13 or newer and the UV package manager from Astra. Install UV with one of these options.

Option A: Install UV directly with Python package manager.

pip install uv

Install in Claude Desktop

Install Claude Desktop and then install the MCP packager, build the extension, and install it in Claude Desktop.

npm install -g @anthropic-ai/mcpb

Clone the Windows-MCP repository and build the Desktop Extension MCPB.

git clone https://github.com/CursorTouch/Windows-MCP.git

cd Windows-MCP

npx @anthropic-ai/mcpb pack

Open Claude Desktop and install the extension through Settings > Extensions > Advance Settings > Install Extension, selecting the .mcpb file.

Install in Perplexity Desktop

Install Perplexity Desktop, clone the Windows-MCP repository, then configure the connector to run the local MCP server.

git clone https://github.com/CursorTouch/Windows-MCP.git

cd Windows-MCP

In Perplexity Desktop, go to Settings > Connectors > Add Connector > Advanced. Use the following config to run the local server.

{
  "name": "windows-mcp",
  "command": "uv",
  "args": [
    "--directory",
    "<path to the windows-mcp directory>",
    "run",
    "main.py"
  ]
}

Install in Gemini CLI

Install Gemini CLI, clone the repository, and add the MCP config to Gemini’s settings so it can start the local server.

npm install -g @google/gemini-cli

Edit Gemini settings to include the Windows-MCP config.

{
  "theme": "Default",
  //MCP Server Config
  "mcpServers": {
    "windows-mcp": {
      "command": "uv",
      "args": [
        "--directory",
        "<path to the windows-mcp directory>",
        "run",
        "main.py"
      ]
    }
  }
}

Install in Qwen Code

Install Qwen Code, clone the repository, and configure settings to run the local MCP server.

npm install -g @qwen-code/qwen-code@latest

In Qwen Code settings, add Windows-MCP as shown.

{
  //MCP Server Config
  "mcpServers": {
    "windows-mcp": {
      "command": "uv",
      "args": [
        "--directory",
        "<path to the windows-mcp directory>",
        "run",
        "main.py"
      ]
    }
  }
}

Install in Codex CLI

Install Codex CLI, clone the repository, and configure the Codex config to run the local MCP server.

npm install -g @openai/codex

Add the Windows-MCP config to Codex config.toml.

[mcp_servers.windows-mcp]
command="uv"
args=[
  "--directory",
  "<path to the windows-mcp directory>",
  "run",
  "main.py"
]

Available tools

Click-Tool

Click on screen at specified coordinates to interact with UI elements.

Type-Tool

Type text into a focused element, with optional clearing of existing text.

Scroll-Tool

Scroll vertically or horizontally within a window or region.

Drag-Tool

Drag the cursor from one point to another.

Move-Tool

Move the mouse pointer to a target location.

Shortcut-Tool

Simulate pressing keyboard shortcuts like Ctrl+C or Alt+Tab.

Wait-Tool

Pause execution for a defined duration.

State-Tool

Capture a snapshot of the system state including language, active apps, and UI components.

App-Tool

Launch, resize, and switch between applications from the start menu.

Shell-Tool

Execute PowerShell commands.

Scrape-Tool

Scrape information from webpages or UI elements.