agent-browser

Automate browser-based workflows using LLMs and computer vision to navigate complex websites without custom scripts.

Agent-browser technology (exemplified by Skyvern) replaces brittle, DOM-dependent scrapers with AI agents that perceive web interfaces like humans do. By leveraging Playwright and computer vision, these agents handle dynamic elements, multi-step forms, and unexpected pop-ups across any URL. For example, an operator can automate insurance claims processing or flight bookings by providing high-level goals rather than writing 500 lines of CSS selectors. It effectively turns the open web into a structured API for high-volume data extraction and task execution.

https://github.com/skyvern-ai/skyvern