AI Agent Test is a module that allows you to test the AI Agents decision making process over and over again, using a known input context against any provider and model combination.
It is very easy to be tricked when you use the Agent Explorer on some random samples of data to think that you have created a production ready agent, but the real world is much more complex. If you develop an agent you have domain knowledge about how to instruct it correctly, and you inputs will be very biased with correct vocabularies and context as compared to the real world.
The AI Agent Test module allows you to create a test suite of actual prompt that testers/end-users did that can be run against any agent, so you can be sure that it will work in production.
To read more about how it works, check the AI Documentation page about AI Agents Test