Skip to main content
from askui import VisionAgent

with VisionAgent() as agent:
    agent.act("Click 'Sign in', fill email with 'test@example.com', submit")
No selectors. No element IDs. Natural language instructions.

Capabilities

FeatureDescription
Vision-basedSees pixels, not DOM. Works on any UI.
Cross-platformDesktop, web, Android, iOS, HMI
Background modeRuns without taking over your screen (Windows)
Natural languageDescribe tasks in plain English

Use Cases

  • Agentic development: Give Claude Code/Cursor UI control
  • Test automation: CSV-based test suites with auto-reporting
  • Process automation: Automate workflows across any app