Stop waiting 3 seconds for screenshots. Drive Windows at the speed of native code.
Directly traverses the Win32 UI Automation (UIA) accessibility tree to click buttons, inspect window states, and input text in sub-30 milliseconds without heavyweight computer vision models.
| Automation Paradigm | Action Latency | Token / Model Cost | Click Reliability |
|---|---|---|---|
| Vision-Based Agent (Screenshot + VLM) | 2,800ms - 4,500ms | High (Image Tokens) | Brittle (Resolution Dependent) |
| Auto_OS Win32 UIA Tree | <28ms | Zero Model Tokens | 100% Deterministic Handle |
from auto_os import DesktopActuator
actuator = DesktopActuator()
# Find element by Win32 AutomationID in <10ms
button = actuator.find_element(name="Submit", control_type="Button")
button.click()