A few months ago, I had this annoying habit of opening my editor and just... hoping for the best.
I’d use Claude for one thing, ChatGPT for another, maybe a quick code review here, a rewrite there. Nothing felt expensive in the moment. A few bucks. A few prompts. No big deal.
Then one day I looked up and realized I had no actual idea what I was spending.
Not “roughly.” Not “probably okay.” I mean I genuinely couldn’t answer a basic question:
What did AI cost me this week?
That’s the moment TokenBar started making sense to me.
I didn’t want a dashboard. I didn’t want another tab. I didn’t want to go hunting through billing pages after the damage was already done.
I wanted the number in the one place I actually look all day: my Mac menu bar.
So I built TokenBar.
It’s tiny on purpose. It sits there and shows me what I’m using in real time, so I can catch myself before I turn a “quick ask” into a surprise bill. That sounds almost stupidly simple, but honestly that’s the point. Most cost problems in AI feel invisible until they’re not.
The thing that changed for me wasn’t just the number. It was the behavior.
When I can see token usage while I’m working, I write more carefully. I stop blasting huge context dumps everywhere. I’m way less casual about asking for giant rewrites when I only need one paragraph fixed. I still use AI heavily — I just use it with my eyes open now.
That’s what I wish I’d had earlier: not a spreadsheet, not a monthly postmortem, just a live signal.
If you’re building with AI and you’ve had that weird feeling that your costs are creeping up without you really noticing, that was me too. I built TokenBar because I got tired of treating spending like something I’d deal with later.
Later is how you get burned.
Now the number lives where I can actually see it.
More here if you want it: tokenbar.site
Top comments (0)