clawdcursor for Openclaw

A cross-platform automation server that gives AI agents the hands to control any desktop application via REST or MCP.

amrdab
v0.7.5
Apr 5, 2026
0
864
0

Install & Download

1. ClawHub CLI

The fastest way to install a skill directly from the registry.

npx clawhub@latest install clawdcursor

2. Manual Installation

Copy the skill folder to one of these locations

Global
~/.openclaw/skills/
Workspace
<project>/skills/

Priority: Workspace > Local > Bundled

3. Prompt Installation

Copy this prompt to OpenClaw to install it automatically.

Help me install clawdcursor using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).

Prefer to download?

Get the raw skill files in a ZIP archive.

What is clawdcursor?

ClawdCursor is a powerful OS-level desktop automation tool server designed to bridge the gap between AI reasoning and physical execution. It provides a robust collection of 42 tools compatible with any AI model capable of function calling, such as Claude, GPT, or Gemini. As a core part of the Openclaw Skills ecosystem, it operates as a local server, ensuring privacy while enabling complex interactions across Windows, macOS, and Linux.

Unlike traditional automation frameworks, ClawdCursor is model-agnostic and functions in three distinct modes: a REST API server, an MCP (Model Context Protocol) host, or an autonomous agent pipeline. It is specifically designed for the last mile of automation where native APIs or CLIs are unavailable, allowing agents to interact with GUI elements, manage windows, and manipulate browsers through the Chrome DevTools Protocol.

clawdcursor Use Cases

  • Automating tasks in legacy desktop software lacking modern APIs.
  • Deep browser automation using CDP for complex form filling and data extraction.
  • Facilitating cross-application workflows like copying data from a CRM to a local spreadsheet.
  • Deploying autonomous agents to handle multi-step desktop processes with safety-gated confirmations.

How clawdcursor Works

  1. Orient: The skill uses read_screen or get_windows to analyze the accessibility tree and identify interactive elements.
  2. Act: It executes specific actions such as smart_click, key_press, or type_text based on the agent's logic.
  3. Verify: Every action is verified by checking tool return values, window states, or visual changes to ensure success.
  4. Repeat: The loop continues until the task is complete or the agent reaches a terminal state.

clawdcursor Setup

To begin using this skill, install it via npm and initialize the environment:

npm install -g clawdcursor
clawdcursor consent --accept
clawdcursor serve

For MCP integration, add the server to your configuration file:

{
  "mcpServers": {
    "clawdcursor": {
      "command": "clawdcursor",
      "args": ["mcp"]
    }
  }
}

clawdcursor Data Schema & Taxonomy

ClawdCursor organizes data through a standardized toolset and coordinate system:

Data Type Description
Coordinates 1280px-wide viewport image-space coordinates; DPI scaling is handled automatically.
Authentication Mutating requests require a Bearer token stored in ~/.clawdcursor/token.
Metadata Skill configuration and workspace logs are stored in ~/.openclaw/workspace/skills/clawdcursor.
Perception Uses Accessibility trees (UIA, JXA, AT-SPI) and OCR for structured screen data.

clawdcursor Advanced Features

  • Adaptive Learning: Use the POST /learn endpoint to teach the agent specific UI patterns and shortcuts for better efficiency.
  • CDP Integration: Native support for Chrome DevTools Protocol allows for direct DOM interaction and JavaScript execution.
  • Autonomous Delegation: Pass complex strings to delegate_to_agent for full end-to-end task execution within the Openclaw Skills framework.
  • Multi-Tier Safety: Built-in safety levels (Auto, Preview, Confirm) ensure high-risk actions like deletions require explicit user approval.

SKILL.md


Loading

Related Openclaw Skills

METADATA

Github Stars: 0
forks: 0

Featured*