You're paying OpenAI/Anthropic per token. Your customers are using your app. Somewhere in between, there's a spreadsheet, manual invoices, and zero idea who's burning through your budget.
I have seen companies keep getting hit with surprise LLM bills, retry loops, users spamming AI features, and no way to know which customer caused a cost spike. So I started building TokenTrail to fix it for myself and figured others might have the same problem.
What it does:
Set hard budget limits that kill requests at $X (no more Monday morning surprises, though customisable from the user side)
Track cost per customer, not just total API spend
See what each step in a multi-agent workflow actually costs
Basic smart routing to cheaper models for simple queries
