LaunchKit · 2026
Back to Skills

computer-use

Full desktop computer use for headless Linux servers and VPS.

30
10.7k downloads
by @ram-raghav-s

Setup & Installation

openclaw skills install @ram-raghav-s/computer-use

Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:

npx clawhub install computer-use

Version History

v1.2.1Feb 15, 2026 - Version note by skill developer:

- Improved description to highlight X11-level automation for undetectable interactions (vs browser tools). - Clarified Chrome usage: added instructions to check user namespace support and avoid unnecessary --no-sandbox flag (improves browser stability). - Updated manual setup instructions to suggest using the setup script for generating service files. - Various documentation refinements for clarity and accuracy.

What This Skill Does

Full desktop GUI control for headless Linux servers via a virtual display (Xvfb + XFCE). Provides 17 actions, including clicks, typing, scrolling, and screenshots, executed through xdotool at the X11 level. Live desktop viewing is available through a browser-based noVNC client or any VNC client.

Operates at the X11 level rather than the browser API level, so websites cannot detect the automation.

When to use it

  • Automating GUI desktop apps on a headless VPS
  • Running browser sessions that bypass automation detection
  • Watching and debugging remote desktop activity in real time
  • Scripting multi-step GUI workflows on a cloud server without a monitor
  • Testing desktop software in a CI environment with no physical display

Example Workflow

Here's how your AI assistant might use this skill in practice.

INPUT

User asks: open Chrome on the remote server, navigate to a URL, and confirm the page loaded

AGENT
  1. 1Take a screenshot to see the current screen state
  2. 2Launch Google Chrome with the appropriate sandbox flags for the kernel
  3. 3Type the target URL into the address bar
  4. 4Press Return to navigate
  5. 5Take a screenshot to verify the page loaded
OUTPUT

Base64-encoded PNG screenshot of the loaded page on the remote desktop