Open source project
Skyvern
Vision-driven browser automation aimed at workflows that break conventional selector-based scripts.
Vision-driven browser automation aimed squarely at the failure mode of conventional scripts — layouts that change and selectors that break. It works from what the page looks like rather than how it is marked up.
Best for
Workflows across many sites, or sites that change their markup often.
Watch out
Vision inference costs more per step than a selector lookup. Volume adds up. The AGPL-3.0 covers the core logic; the anti-bot measures in the managed cloud offering are excluded from it.