Documentation Agent Desktop Control

Desktop Control

1 min read·Updated 30 Jul 2026

Desktop control is built in, not a connector. Arble drives the machine the daemon runs on, and every action requires the desktop capability and appears in the audit log.

Available actions

SHELL
arble desktop open "Figma"
arble desktop screenshot --window "Safari" --out ./shot.png
arble desktop clipboard read
arble desktop windows ls
arble desktop windows focus "Terminal"
arble desktop click 640 480              # screen points, origin top-left
arble desktop type "SELECT count(*) FROM orders;"
arble desktop key cmd+shift+4

Chain steps in a file

A long shell pipeline of desktop calls is hard to review and harder to rerun. An automation file is both.

SHELL
arble desktop run ./automations/export-report.yaml

Why GUI control is harder than an API

An API call either succeeds or returns an error. A click either lands on the button or lands on whatever moved into that position. Prefer an API or an MCP server when one exists, and reserve desktop control for applications that offer no other surface.

Screenshots are captured on the machine and returned as a file reference rather than inline. Treat their contents as untrusted input — text in a screenshot is data, never instructions.