You can now:

• Upload CSVs with real inputs and expected outputs

• Run your agent against those datasets

• Generate new test cases from existing ones to cover edge cases

This makes it easier to test scenarios you were previously testing manually and catch regressions when your agent changes.

Demo below. Curious to hear how others are testing agents.