| name | Windows MCP Overview |
| description | This skill should be used when the user asks "what Windows MCP tools are available", "how to automate Windows", "which Windows tool should I use", "Windows MCP overview", "list Windows MCP capabilities", "what can Windows MCP do", "get started with Windows automation", or needs guidance on selecting the right Windows-MCP tool for a task. |
| version | 1.0.0 |
Windows MCP Overview
The Windows-MCP server (by CursorTouch) provides 16 tools for comprehensive Windows desktop automation. This skill covers tool selection, capabilities, and high-level usage patterns.
Prerequisites
- Python 3.13+ installed
- UV package manager (
pip install uv)
- Windows 7/8/8.1/10/11
- English as the default Windows display language (or disable the App tool for other languages)
Tool Categories at a Glance
Screen Interaction
| Tool | Purpose |
|---|
| Click | Click at screen coordinates (left/right/double) |
| Type | Input text into focused elements with optional clear |
| Scroll | Vertical/horizontal scrolling at specific coordinates |
| Move | Mouse movement with optional drag support |
| Shortcut | Execute keyboard shortcuts (Ctrl+C, Alt+Tab, etc.) |
| Wait | Pause execution for a specified duration |
Visual Capture
| Tool | Purpose |
|---|
| Screenshot | Fast desktop capture with cursor and active window info |
| Snapshot | Full state capture with interactive element IDs and DOM extraction |
Application Management
| Tool | Purpose |
|---|
| App | Launch, resize, move, switch, minimize, maximize, close windows |
System Operations
| Tool | Purpose |
|---|
| Shell | Execute PowerShell commands |
| Process | List running processes or terminate by PID/name |
| Registry | Read, write, delete, or list Windows Registry keys/values |
| Notification | Send Windows toast notifications |
Data Operations
| Tool | Purpose |
|---|
| Scrape | Extract information from web pages |
| Clipboard | Read or write Windows clipboard content |
| MultiSelect | Select multiple items (files, folders, checkboxes) |
| MultiEdit | Input text into multiple form fields simultaneously |
Tool Selection Decision Tree
Need to see the screen?
- Quick visual check ->
Screenshot
- Need interactive element IDs ->
Snapshot
- Need web page DOM content ->
Snapshot with use_dom=True
Need to interact with GUI elements?
- Click something ->
Snapshot first to locate, then Click
- Enter text ->
Type (with clear: true to replace existing text)
- Scroll a page/list ->
Scroll with target coordinates
- Use a keyboard combo ->
Shortcut
- Drag an item ->
Move with drag: true
Need to manage applications?
- Open a program ->
App (launch)
- Arrange windows ->
App (resize/move)
- Switch focus ->
App (switch) or Shortcut (Alt+Tab)
- Close a program ->
App (close) or Shortcut (Alt+F4)
Need system-level operations?
- Run a command ->
Shell
- Monitor/kill processes ->
Process
- Modify system settings ->
Registry
- Alert the user ->
Notification
Need data operations?
- Extract web content ->
Scrape
- Transfer data via clipboard ->
Clipboard
- Select multiple items ->
MultiSelect
- Fill multiple fields ->
MultiEdit
Core Automation Workflow
Most Windows automation follows this cycle:
- Observe — Capture screen state with
Screenshot or Snapshot
- Identify — Locate target elements using
Snapshot element IDs or coordinates
- Act — Perform the action (
Click, Type, Shortcut, etc.)
- Verify — Capture screen again to confirm the action succeeded
- Wait — Use
Wait between steps if the UI needs time to update
Performance Notes
- Typical action latency: 0.2-0.9 seconds
Screenshot is significantly faster than Snapshot
Snapshot with use_dom=True is the slowest but provides the most data
- Batch tools (
MultiSelect, MultiEdit) outperform sequential individual actions
- Prefer
Shell over GUI interaction for any task that can be scripted
Additional Resources
Reference Files
For complete parameter documentation and examples for every tool:
references/tool-reference.md — Detailed parameter reference for all 16 Windows MCP tools with usage examples and tips