Skip to main content
MCP server CX Codex

clawdcursor — Claude Code MCP server: desktop control for agents

Lets an agent drive the desktop: turns the screen into a UI map and clicks elements by stable id. Useful when the app has no API but the task still has to get done.

Install into Claude Code · Platform CLI All methods and platforms ↓
claude mcp add clawdcursor -- npx -y clawdcursor
Open on GitHub

What clawdcursor does

clawdcursor sits between your agent and the operating system. It reads the screen through the accessibility tree, builds a map of the interface, and acts on elements by their identifiers. Screenshots and vision come last, only when the markup can't be read.

Every action is checked afterwards, and the result goes back to the agent. One safety checkpoint gates the whole flow. If a sign-in, a 2FA code or an upgrade page blocks the way, the run pauses: the screen edge turns blue, the agent asks you for that single step, then carries on from where it stopped.

You need Node.js 20 or newer. On macOS, refusing screen-recording permission breaks only screenshots and OCR — windows and keystrokes keep working. Windows, macOS and Linux are all supported, so test on the system where the target app actually runs.

What it helps with

  • run an operation in an app that has no public API
  • fill a desktop form field by its name
  • pause the agent at a login and hand that step to a person
  • confirm the expected result showed up after a click
  • read a page's text without touching the clipboard

Example prompts for the agent

  • Open the app, find the field called User code and type the value from the email
  • Fill in the form, but stop at the sign-in screen and ask me to log in
  • Copy the full text of the open document and summarise it

How to install clawdcursor

Platform CLI

claude mcp add clawdcursor -- npx -y clawdcursor

Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more

Questions about clawdcursor

How do I install clawdcursor in Claude Code?

Run claude mcp add clawdcursor -- npx -y clawdcursor.

What does clawdcursor do?

It operates the desktop on the agent's behalf: locates UI elements by name or role, clicks and types, then verifies the outcome. Screenshots and vision kick in only when the markup can't be parsed.

What does clawdcursor need to run?

Node.js 20 or newer. On macOS, screenshots and OCR also need screen-recording permission; without it those two features fail while the rest keeps working.

Similar MCP servers

All MCP servers for Agent setup: memory, context, orchestration →

Project sponsors

Companies that keep this analytics open
Gold sponsor

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor