Token maxing.

Trying to dial in sub agents. Have special models read only over a "bridge" and the usual models for edits. Astra low is the orchestrator. So far it can go continuously 48 hrs before weekly limit runs out. Work does not suffer. Pro 20x. Thoughts?