ClawX Computer Use skill

An agent fills fields and selects menu items in a local application, checking the screen before and after each action.

  • 7,601 ★ on the repository
  • Last commit
  • MIT
Skill

Broad access

Why this level

  • The agent gains broad local UI, input and window-content access; screenshots may reach a model and actions can affect accounts and files.
All reasons and checks

valuecell-ai/clawx

Install

Works in

In your terminal, with SkillFoxx CLI

npx skillfoxx add skills/clawx-computer-use

Detects the agents on your machine, shows the access level and pins the version.

Other ways to install

Access and checks

How we review

Broad accessThe tool can change data, run commands or handle secrets.

  • The agent gains broad local UI, input and window-content access; screenshots may reach a model and actions can affect accounts and files.

This is third-party code. Review the repository files before installing.

What it does

ClawX bundles a `computer-use` skill and a native CUA CLI for controlling applications. The skill guides an agent to inspect the window, locate an element, act and observe the result. It suits tasks in applications without an API available to the agent. The user enables Computer Use on its settings page and grants OS permissions. The skill instructs the agent to request confirmation before consequential external actions. If a click or text entry has an uncertain result, the agent must inspect the window again before repeating it.

Who it is for. ClawX users on macOS and Windows delegating form entry and other desktop application tasks to an agent.

Good fit when

  • An application lacks a suitable command and the user chooses UI interaction.
  • An agent compares window state before and after an approved action.
  • A task requires menu navigation or an agreed field entry.
  • After a connection interruption, the agent inspects the window again to avoid clicking a button twice.

Not a fit when

  • Computer Use would run on Linux or another unsupported platform.
  • The user has not enabled it or granted OS permissions.
  • The task requires automatic handling of login, MFA or OS permission dialogs.

Example request

Inspect the selected app window and locate the required field; stop before form submission and show the completed values.

Limitations

Computer Use supports macOS 13 or later on Intel and Apple processors and Windows 10 or later on x64; this feature is unavailable on Linux. Both the model and its provider endpoint must accept images. At startup, ClawX replaces a modified `~/.openclaw/skills/computer-use` directory with its bundled copy; keep custom variants under a different name. Disabling the feature stops the driver but preserves existing OS permission grants.

How to disable. Turn off Computer Use on its ClawX page. Separately revoke OS control and screen-capture permissions if no longer needed.

FAQ

Does selecting the skill enable computer control?

No. The user enables the feature separately.

Where do I enable Computer Use?

Open Settings, then Advanced, and enable Developer Mode. Open Computer Use in the sidebar and enable the feature on that page.

Why is a screenshot saved but invisible to the agent?

The model or provider endpoint may accept only text. Select an image-capable model and stop blind input until the agent can view the screen.

Can this integration record video?

No. Recording is unsupported in its one-shot CLI workflow.

Does cancelling a command undo a click?

No. An action already accepted by the driver may have occurred.

Discussion

No questions or reviews yet.

Skills

An MCP server that lets an agent drive windows and apps: it reads the screen as structure and acts by element name

MCP serverBroad access408Repository stars

A skill that wires Agent-S into OpenClaw: the agent hands over a task in words and Agent-S clicks through desktop windows itself.

SkillBroad access13KRepository stars

A local macOS CLI that lets an agent read the screen via OCR, click and type, with no cloud and no vision model

CLIBroad access71Repository stars

Open Computer Use

open-computer-use

An open MCP server for desktop control: the agent reads app state and clicks, types and scrolls on macOS, Linux and Windows

MCP serverBroad access2.9KRepository stars

Mano skill

Mano skill (mano-cua)

A CLI and skill for desktop control from a text task: screenshots, clicks, typing and scrolling, in the cloud or locally on Apple Silicon

SkillBroad access47Repository stars
Official

computer-use-linux is an MCP server with an on-screen agent activity indicator that lets user clicks reach the application.

MCP serverBroad access867Repository stars

README in short

Screenshots are saved in the task workspace but may reach the model provider when read. The skill is not a system sandbox or emergency-stop guarantee.

README badge

Report a problem