Ditto - would be great to get some clarity on credit usage. We've also run into the issue of the Agent Builder going in loops and having to resort to manually fixing its issue (which is understandable. AI needs exact prompting, hand holding etc) but the biggest issue is the actual agents themselves, the ones we've built, quitting unexpectedly and then having to re prompt it to continue where it left off. Each response we get back from it costs credits, and it responds numerous times just informing the user that it's going to do what we ask, wasting large numbers of credits.
