The tools that tell you what AI is costing you

The problem cost-tracking tools solve is not measurement — every provider already meters you. The problem is that the measurements don't add up. One provider counts tokens, another GPU-hours, a third characters, a fourth requests. Each lives behind its own login and bills on its own cycle. No single page anywhere answers the only question that matters at month-end: what is AI costing me, in total, and where is it heading?

There are really only a few ways to close that gap, and they trade off along the same axis: how much of your own work each one asks for. A row of provider dashboards is free but manual — you read each console and add it up by hand. A spreadsheet captures the total but has to be fed. Observability platforms and AI gateways compute cost precisely, but earn it by sitting inside your code or your request path. A billing dashboard reads each provider's own usage API from outside your application and converts it to money, so the total assembles itself with nothing installed.

Which one is right depends on what else you need. If you also want per-request tracing or routing, the tools that live in your code path pay for themselves. If the question is purely what am I spending across my providers this month, the lowest-friction answer is the one that reads the numbers you're already generating and leaves your application untouched.

The guides below walk each approach — what it captures, what it costs you to run, and where a tracked total can drift from the invoice.