Habit combines human operational demonstration with machine precision. Learn how our four core systems execute repetitive computer operations reliably.
Unlike fragile HTML scraping or pixel-based vision that breaks on font changes, Habit anchors to the operating system's Accessibility Tree (UI Automation on Windows, NSAccessibility on macOS). It identifies buttons and inputs by semantic role.
When you teach Habit, it captures both mouse/keyboard actions and your ambient voice narration. The voice clarifies your intent: "I check if this row has an exception, then query NetSuite for this PO." Habit compiles this into structured rules without code.
Execution runs locally via a lightweight Tauri (Rust) and Python sidecar. Before clicking, Habit verifies application state, enters data at a human cadence, and stops immediately whenever an unexpected condition appears.
Habit never uses language models to calculate invoice totals, deduction subtotals, or quantities. All tabular calculations and file transformations execute via deterministic native code with 100% mathematical accuracy.