So for that we use

, we run Agents on n machines and we put those tasks to queues and tasks are run on n machines.
Do I understand that correctly?

Yes, indeed. You can have all of these agents listen to a single queue where the HPO will place all experiments