I Stopped Hiding My AI Spend in a Dashboard

Last winter I did the thing every solo dev does when they want to feel in control: I opened a dashboard.

It had charts. It had filters. It had all the right vibes. But it also had the one thing I actually needed to know buried under three clicks and a tab I only opened after I had already spent the money.

So I kept missing the moment.

I wasn’t building a big SaaS team. I was just trying to ship TokenBar, a tiny menu bar app for watching LLM usage and cost in real time. That’s the whole point: if the bill is invisible until the end of the month, you’re already too late.

What changed for me was embarrassingly simple. I stopped treating AI spend like a report and started treating it like a live signal.

That meant putting the number where I’d actually see it: right in the menu bar.

Not because dashboards are bad. They’re fine if you’re doing finance cleanup on a Saturday. But during actual work, I need frictionless. I need to glance up and instantly know whether the last hour of coding was efficient or stupid.

That’s the mindset behind tokenbar.site. I built it for people like me who live inside their editor all day and don’t want to context-switch just to check whether their quick experiment turned into a $14 afternoon.

The first time I saw live token usage while I was coding, I immediately changed how I worked. I trimmed prompts. I stopped dumping giant files into every request. I became way less casual about model choice.

Not because I became frugal in some noble way.

Because seeing the cost in real time made it feel real.

And honestly, that’s the part nobody talks about enough. AI spend isn’t just a billing problem. It’s a feedback problem. If the feedback arrives two weeks late, you can’t fix the behavior that caused it.

So yeah, I built a menu bar app instead of another dashboard.

Because I didn’t need another place to look.

I needed the number in my face while I was still deciding what to do next.