How do you work with multiple agents?

We train with a single agent. is that the question?