Finding the loop that burned an AI agent’s token budget
In live AI agent systems, the bigger cost problem may not be steady usage. It can be the agent repeating the same failed action, changing its plan a little each time, until the is gone. A global spending limit can stop the loss, but it does not show which branch, call, or tool used the money.
Teams running agents need clearer cost tracking for each call and each tool. They also need rules for when to stop, such as a hard spending cap, a no-progress limit, or both.
Key points
- AI agents can waste tokens by retrying the same failed action in slightly different ways.
- A global spending cap stops runaway cost but does not explain where the money went.
- Per-call and per-tool cost tracking can reveal the expensive part of a failed run.
- Stopping rules should consider both hard cost limits and lack of progress.
- Retry loops are a real risk for agent cost control.