I did not build TokenBar because I love menu bar widgets.
I built it because I kept getting surprised by AI usage costs.
A prompt here, a few retries there, a bigger model on a task that did not need one, and suddenly the bill looked nothing like the thing I had in my head.
What changed things for me was simple visibility. If I can see what is happening in real time, I make better calls:
- I notice waste faster
- I stop guessing about usage
- I can choose a cheaper model when the task is simple
- I feel less blind when the bill arrives
TokenBar is my attempt to make token usage feel like something you can actually watch instead of something you discover later in an invoice.
I am still a solo dev shipping this in public, and the main lesson so far is that small feedback loops beat heroics.
If you are building with AI and getting surprised by costs, the problem may not be the invoice. It may be the lack of visibility.
Top comments (0)