Unit-testing AI Agents in the Browser - Browserbase - Headless Web Browser API

Instead of trying to evaluate every aspect of an AI agent, we can focus on a simpler question: "Given a specific instruction like "click the button", was the expected button clicked?"