computer-use:take_screenshotLABSCO SUMMARY
Windows automation without a browser in the way.
Screenshot, find the window, click and type: the loop browser drivers cannot reach, because this one works on the desktop itself through the Windows API.
What it is
A self-contained C# server for Windows that calls the Windows API directly. It captures screens across multiple monitors, enumerates and controls windows, and sends mouse and keyboard input; a macro tool replays a whole sequence of those actions in one call. It ships as a single executable.
What you get
- Screenshots with flexible targeting — a monitor, a multi-monitor desktop, or one specific window (computer-use:take_screenshot)
- Window management: list every visible window, read the active one, focus by ID or name, and minimise, maximise or restore (computer-use:list_windows, computer-use:get_active_window, computer-use:focus_window, computer-use:manage_window)
- Input: click with a choice of button, move the pointer, scroll at a coordinate, type Unicode text and press keys with modifiers (computer-use:mouse_click, computer-use:mouse_move, computer-use:scroll, computer-use:type_text, computer-use:press_key)
- Macros: computer-use:run_macro replays a full action sequence with configurable timing and its own error reporting
- Status: computer-use:get_server_capabilities reports what the server can do and its current status
Requirements
Windows 10 or newer. Download MCPComputerUse.exe, place it where you want and point your client's command at it — it is self-contained, with no runtime to install. It can click and type anywhere on that desktop, so run it on a machine where that is acceptable.
Setup effort
Follow the published steps — Download the latest release from GitHub
USE CASES
01
Working with an application that has no API
Capture the screen, focus the window, then click and type — the desktop equivalent of driving a browser.
02
Capturing what is actually on screen
computer-use:take_screenshot targets a single monitor, the whole multi-monitor desktop, or one window.
03
Repeating a fiddly sequence reliably
computer-use:run_macro replays the whole sequence with its own timing instead of a dozen separate calls.
04
Finding out what the machine is doing
Listing visible windows and reading the active one answers where input would land before you send any.
12 tools
Listed in project docs. Your client may show a different list.
computer-use:list_windowscomputer-use:get_active_windowcomputer-use:focus_windowcomputer-use:manage_windowcomputer-use:mouse_clickcomputer-use:mouse_movecomputer-use:type_textcomputer-use:press_keycomputer-use:scrollcomputer-use:run_macrocomputer-use:get_server_capabilitiesBEFORE YOU INSTALL
What to know first
Tools you get
12 tools named in the project's documentation. The computer-use: prefix comes from the server key in your config. OCR, image matching and macro recording do not exist yet.
Requires
Windows 10 or Windows 11
Where it runs
The documented client runs the server on your machine.
INSTALLATION
Connect your client
Server setup
Download the latest release from GitHub
Extract MCPComputerUse.exe to your desired location
No installation required - single executable file
Add the mcpServers block below to your Claude Desktop configuration
MCP Clients
Claude Desktop
CLAUDE DESKTOP · JSON
{
"mcpServers": {
"computer-use": {
"command": "<YOUR_PATH_TO_MCP_COMPUTER_USE>",
"args": []
}
}
}
