Token Spend Review
A daily review of how I spent my Claude, Codex and Gemini allowances: which model did which job, on which project. It refreshes a private dashboard and suggests ways to spend less without lowering quality.
- Why
- I run several AI subscriptions and a crew of agents, and I couldn't tell where my limits went. I wanted savings that never quietly downgrade the model doing my planning or review.
- How long
- Set up in October 2026
How I built it
- I decided what to measure first: allowance used, led by each subscription's 5-hour and weekly meters, then tokens.
- I had the AI write a script that reads usage from my Mac's own logs and breaks it down by model, job type and project.
- Anything it can't measure shows as 'not measured' with the reason, never left out.
- It suggests changes as numbered options, and changes nothing until I say yes to that one option.
- To copy it you need
- Claude Code, or the AI of your choice, with access to your own usage logs. Gotcha: write down which jobs must keep the strong model before you ask for savings, or the AI will save money by quietly downgrading your best work.
Try it yourself
Copy this prompt and paste it into ChatGPT or Claude to try it yourself.
I pay for more than one AI subscription and I can't tell where my limits go. Korab Eland built a Token Spend Review (on whatarewebuilding.ai) that reads usage from the computer's own logs, breaks it down by model, job and project, refreshes a private dashboard, and suggests savings as numbered options, changing nothing without a yes. Korab's rule: never save money by moving planning or review to a weaker model. Help me build a simple version. First ask which AI tools I use and where they keep usage records. Then make a one-page daily summary, and mark anything you can't measure as 'not measured'. Ask me for anything you need.
Tools I used for this project
