LLM providers give you account-level limits. TokenFence gives you per-workflow limits. Set a dollar cap, pick what happens when it's hit (kill or downgrade model), add two lines of code. Works with OpenAI, Anthropic, and any provider.