Screenshot-driven desktop automation
Linux Computer Use
Give coding agents a controlled screenshot–act–verify loop for Linux desktop software.
Overview
A portable Codex skill and deterministic driver for inspecting, operating, and visually verifying Linux X11 or XWayland desktop applications locally or through a Parallels VM.
Linux Computer Use supplies the mechanical desktop-control layer for a coding agent while keeping visual reasoning and safety explicit. Each meaningful action begins from a known window and screenshot and ends with new visible evidence.
Local X11 and XWayland operation
Parallels-hosted Linux VM backend
Screenshots, windows, focus, keyboard, and mouse
Screenshot-backed smoke tests and UI QA
Built for
Where it fits.
- Coding-agent users testing Linux desktop applications
- Cross-platform UI teams without native Computer Use integration
- Maintainers reproducing and visually verifying GUI behavior in VMs
Compatibility & status
Local control
Application windows
Limited input control
macOS host backend
In the work
Real output,
not a mockup.
Architecture
Clear layers.
Replaceable seams.
Target
The driver selects a local graphical session or a named Parallels Linux guest.
Observe
Screenshots, visible windows, active-window state, and geometry establish the current desktop.
Act
Focus, keyboard, pointer, click, and scroll commands perform one deterministic interaction at a time.
Verify
A follow-up screenshot proves visible state instead of treating a successful process exit as UI evidence.
Quick start
From package
to first frame.
Choose the package workflow used by your repository, then start from the smallest working example. Replace VERSION with the current stable or prerelease version shown on NuGet.
git clone https://github.com/wieslawsoltes/LinuxComputerUse.gitcd LinuxComputerUse./install.sh$linux-computer-use Launch the application in my Linux VM,exercise its primary workflow, and visually verify the result.Package surface
Start focused.
Compose as needed.
Source, docs, samples, and releases
The repository is the source of truth.
Use the current documentation and samples, inspect changes, report issues, or support continued open-source maintenance.