I Started Checking My AI Spend Like a Pulse, Not a Receipt

A few months ago, I thought I had a pretty normal relationship with AI tools.

I’d open Claude, ask it to help me move faster, pay the bill at the end of the month, and move on.

That lasted exactly until I had one of those moments where the number was small enough to ignore, but annoying enough to stick in my head. Not catastrophic. Just enough to make me realize I was treating AI spend like a receipt from a nice dinner instead of a live input to my work.

That was the real problem.

The bill was never the hard part. The hard part was the silence before it showed up.

If I’m being honest, I don’t want another dashboard telling me what already happened. I want a warning light. Something that lives where I work, not somewhere I remember to check once I’m already over the edge.

So I built TokenBar.

It sits in my menu bar and keeps AI usage visible while I’m actually coding. Not after the fact. Not in a spreadsheet. Right there, like battery life or Wi‑Fi. If I’m burning through tokens too fast, I see it while I’m still in the flow and can make a smarter call.

That changed how I work more than I expected.

I’m more careful about when I paste giant context blocks. I notice when I’m about to ask an AI to do something I could solve myself in two minutes. And when I do want to go bigger, I can do it on purpose instead of accidentally drifting into a bill I didn’t mean to create.

I built it because AI costs are not a monthly reporting problem. They’re a live attention problem.

That’s also why I made TokenBar public at https://tokenbar.site — because I think a lot more solo devs are going to hit this exact wall once AI stops being a novelty and starts becoming part of the daily stack.

I don’t think the future of AI billing is a prettier invoice.

I think it’s making the cost visible at the moment of use.

And honestly, that feels a lot more honest.