If you're like most developers right now, you're probably juggling 3-5 paid AI services. Claude for deep reasoning. GPT for quick lookups. Cursor or Copilot for inline code gen. Maybe Whisper for transcriptions.
Each one has its own rate limits, billing cycles, and usage caps. And most of us have exactly zero visibility into how much we're actually consuming across all of them.
The problem I kept running into
I kept hitting rate limits at the worst times. Mid-flow, deep in a refactor, and suddenly Claude tells me I'm done for the next few hours. Or I'd get a surprise bill from OpenAI because I had no idea how much I was burning through their API.
The dashboards exist, sure. But they're scattered across 5+ different websites, each with their own login, their own UI, their own way of showing usage. Nobody checks those proactively. You only find out you're running low when it's already too late.
What I built
I built TokenBar - a macOS menu bar utility that sits quietly in your menu bar and shows your real-time AI token usage across 20+ providers in one glance.
What it tracks:
- Claude (Pro, Team, Enterprise)
- OpenAI / ChatGPT
- Cursor
- GitHub Copilot
- Google Gemini
- OpenRouter
- Perplexity
- Mistral
- And 15+ more
How it works:
- Lives in your macOS menu bar (completely passive, no window to manage)
- Polls your usage from each provider using CLI tokens, OAuth, API keys, and local file probes
- Shows remaining capacity, reset countdowns, and pace indicators
- All processing happens locally on your machine. No telemetry, no cloud sync, no accounts
Key design decisions
Menu bar, not a dashboard. The whole point is that you shouldn't have to go looking for this information. It should be ambient, like a battery indicator. You glance up, you see where you stand, you go back to coding.
Privacy-first. All your credentials stay on your machine. TokenBar never phones home. There's no account to create, no data leaving your laptop.
Broad provider support. Most people don't use just one AI tool. The value is in the unified view across everything you're paying for.
What changed after using it
The biggest shift was behavioral. Once I could see my usage in real time, I started naturally pacing myself. Instead of burning through 80% of my Claude capacity in the morning, I'd spread it out. I stopped getting surprised by rate limits. I started making smarter choices about which AI tool to use for which task.
It's a small thing, but it removed a consistent source of friction from my workflow.
Try it out
TokenBar is $4.99 one-time (no subscription) on the Mac App Store.
Check it out at tokenbar.site
Happy to answer any questions about the technical implementation or the auth approaches for different providers. Some of them were... interesting to figure out.
Top comments (0)