I built a menu bar app to track AI usage limits — here's why
I was juggling Claude, Cursor, Copilot, and OpenRouter — all with different usage limits, reset windows, and credit systems. I kept hitting limits mid-session with no warning.
So I built TokenBar (https://tokenbar.site) — a macOS menu bar app that tracks your AI usage across 20+ providers in real time. It shows limits, credits, reset countdowns, pace indicators, and even detects provider incidents.
Key decisions:
- One-time $4.99 purchase (no subscription)
- Local-first, no cloud dependency
- macOS 14+ native app
- CLI output for automation
The problem is real — I see people on Twitter daily complaining about hitting Claude limits or burning through OpenRouter credits without realizing it. The existing solutions are either provider-specific dashboards (which you have to check manually) or nothing at all.
Currently supports: Claude, Cursor, Copilot, Gemini, OpenRouter, Codex, Augment, Amp, JetBrains AI, Ollama, Warp, Kimi, Kiro, and more.
Would love feedback from anyone else managing multiple AI subscriptions. What's your biggest pain point with tracking usage?
Hey, I'm a CS student building something adjacent to this, a tool that tags AI API spend by feature/workflow so you can see which one is actually expensive, not just total usage against a limit. Since you're already deep in usage tracking, your take would mean a lot.
Quick questions if you don't mind:
No pitch, just trying to validate before I sink time into building the wrong thing. Appreciate any honesty.
This is a useful build because it shows how live AI-spend monitoring has already become. I’m researching one narrow question: for AI products sold through app stores or cloud marketplaces, do teams ever feel a timing mismatch between paying model/API costs continuously and receiving channel payouts weeks later? I can see the cost side clearly, but I’m trying to learn whether the cash-timing side is a real pain or mostly manageable in practice. Curious what you’ve seen from builders around you.