Six months ago I started building three small AI products at the same time. One was a Slack bot that summarized threads. One was a document Q&A tool using RAG. One was a report generator that called Claude every night.
At the end of month two I got an email from my credit card: $412 in API charges. I had no idea which project was responsible.
I logged into OpenAI, Anthropic, and Google dashboards separately. None of them broke costs down by project. I had to grep my logs for request counts, manually look up pricing, and do arithmetic in a spreadsheet I built from scratch at 11pm.
That spreadsheet became the AI API Cost & Budget Tracker.
The core insight: if you are running more than one AI project, you need project-level cost tracking. Your OpenAI bill tells you how much you spent on gpt-4o. It does not tell you that your Slack bot used 60% of it while your RAG pipeline used 30% and your report generator used 10%.
The tracker has five tabs:
- Dashboard: auto-calculates month-to-date spend, projects end-of-month total, flags projects over 80% of budget
- Usage Log: record each API call by project, model, and token count — cost columns calculate automatically
- Pricing Reference: 19 models pre-filled with July 2026 rates across OpenAI, Anthropic, Google, Groq, and Mistral
- Project Budget: set monthly budget per project, track percentage used in real time
- Model Comparison: see cost-per-request for every model at once
After two months I cut my API bill by 34% without degrading any product. My nightly report generator was using gpt-4-turbo for a task gpt-4o-mini handled just as well at one-twentieth the cost.
The tracker is on Gumroad for $17 and Etsy for $14. Includes a 15-page setup guide and a 12-tactic Cost Optimization Playbook.
If you are calling more than one AI provider across more than one project and you do not have per-project cost visibility, this solves that in about fifteen minutes.
Originally published on Medium
Top comments (0)