I run six AI service. Claude for writing and code. OpenAI and Gemini. Ideogram to generate design work. Vectorizer AI to turn that art into clean vector files. Topaz Labs to upscale images. Every one of them charges me a different way, on a different day, through its own login. At the end of the month I had six bills and no single number for what AI was actually costing me. °FAI: The Dashboard That Caught My Runaway API Bills

That’s the trap with usage-based pricing. A software license has one price you can plan around. An API charges you every time you run a job, and the bill only shows up after the money is already spent. I’m not the only one getting surprised. A 2026 Zylo survey found 78% of IT leaders reported AI charges they never budgeted for, and Capgemini found 68% of executives said their company overspent on generative AI, mostly because the cost was scattered and nobody could see it in one place. So I built something to fix it for myself.

What I built

A self-hosted dashboard that pulls all six bills onto one screen. It runs on my own machine, not a paid service. Two of the six, Anthropic and OpenAI, hand over live billing data through their APIs, so those numbers update on their own. The other four, Gemini, Ideogram, Vectorizer, and Topaz, don’t offer that, so I type those costs in by hand, with a reminder bi-weekly to update. The dashboard merges both into a single view. Against my actual output.

How it works

In plain terms: when I open the page and hit refresh, a small server running in the background asks all six services for their numbers at the same time, stacks the live ones on top of my hand-entered figures, and shows the total. The page itself is a single file that runs right in the browser. No install, no setup, no monthly fee to a fourth tool watching my other three.

Each service gets its own card with a monthly budget. The card turns yellow when I cross 80% of that budget and red when I go over. A donut chart breaks down where the money went. A 30-day chart shows the daily trend, so a spike has nowhere to hide.

What the numbers showed

The view paid for itself the first time I looked at it. Vectorizer AI was my single biggest line item, around $64 in March and about $70 in April. Ideogram was next at roughly $61 in March. OpenAI art generation and Gemini were small by comparison. The part that jumped out: Vectorizer’s cost was high against the number of images I actually ran through it. So I pivoted the structure, and started using it a different way.

What this really shows: you can’t control a number you can’t see. The dashboard didn’t cut a dollar on its own. It just put every dollar on one screen. That’s the whole job.

Tools used:

• API Spending Dashboard (self-hosted, tracks all six services in one view)
• Express (the background server that fetches all six bills at once)
• React 18 + Recharts (the in-browser page and its charts, no build step)
• Anthropic + OpenAI billing APIs (live cost data)
• Manual entry (Gemini, Ideogram, Vectorizer AI, Topaz Labs)