AI Costs: The tool that tells you Claude, Gemini spending

Using AI is great, until the bill arrives. Models like Claude or Gemini charge you for every "token" they process, basically a word or a tiny piece of text. The issue is, tracking these costs can become a nightmare, like trying to count grains of sand on a beach.

So, a GitHub user, damejan80, decided to fix this by creating tokentab. This little program was born to bring some transparency to the wallets of those using these services. Finally, a light in the dark world of AI spending.

Tokentab isn't an app with a thousand flashing lights; it's a command-line tool. It grabs the log files from your sessions with Claude Code, Google Gemini CLI, and Codex. Basically, it reads what you asked and what they replied. Then, with a bit of magic, it converts everything into a monetary cost.

This means that instead of waiting for the monthly shock, you can check day by day. You'll need a bit of familiarity with the terminal, but the result is a clear picture of your expenses. Not bad for a program that doesn't even ask for coffee.

Once the data is processed, tokentab spits out a detailed report. You can see how much you've spent with a specific model, for example, "just Claude 3 Opus." Or, if you're working on multiple projects, it shows you the costs broken down by each one. It even tells you how much you burned on a specific day.

The tokentab tool provides detailed cost analysis by model, project, and day. In short, it's like having a personal accountant for your AI, but without the bad jokes. It helps you understand where you might cut back, or at least not fall off your chair when the invoice arrives.