A few weeks ago I got the kind of AI bill that makes your stomach drop.
It wasn't even a crazy business expense. It was just my side project, a couple of coding sessions, a bunch of "quick" experiments that turned into an hour each, and one bad habit I already knew I had: I never checked the meter while I was working.
That is the part nobody talks about.
People love arguing about which model is cheaper, or whether prompt engineering is worth it, or how much AI will save you in time. Fine. All of that matters. But the real trap is that the cost is invisible in the moment. You are in flow, the tab is open, the prompt box is still there, and every extra test feels harmless.
Then the invoice lands and suddenly you're doing math in hindsight.
I built TokenBar because I wanted the number in front of me while I was actually coding. Not in a dashboard. Not in a monthly report. Right there in the menu bar, where I could see it without breaking focus.
That changed my behavior almost immediately.
I started noticing which sessions were cheap and which ones were quietly burning money. I caught myself doing the "one more prompt" loop way earlier. I stopped treating every task like it deserved the fanciest model available. And the weirdest part was how small the fix felt compared to the result.
I didn't need a new budgeting system.
I needed a speedometer.
That is what tokenbar.site is for: a simple macOS menu bar app that shows AI token usage and cost while you work, so you can catch runaway spending before it turns into another bad invoice story.
I'm still a solo dev. I still use AI every day. I just don't want the bill to surprise me again.
Top comments (0)