Gives AI agents a dedicated browser-automation CLI using accessibility-tree references to navigate pages, click elements, fill forms, capture screenshots, and perform repeatable web operations programmatically.