Automate agent evaluation with the Evaluation APIs

When you build an agent in Microsoft Copilot Studio , you want confidence that it behaves exactly as intended: answering correctly, using the right tools, and following the logic you designed. Agent Evaluation (generally available) provides this foundation by allowing you to define test sets, run them against your agent, and understand how it performs. As agents evolve from experimentation into real production scenarios, this foundation becomes part of an ongoing process. Evaluation is no…