Support multiple agents interacting with each other (automatically)

[EwoutH]

To simulate team and agent interactions, it would be great if you could define multiple agents with each a different name/goal, and then start the simulation with a prompt to one (or multiple) agents. From then you can observe which agents decides to contact which and how those interactions look like.

The next step would be instead of all agents being aware of each other existence, initializing them with a list of agent that they know, and allow them to only interact with those agents, until introduces to others.