Stop hitting Claude rate limits mid-session: a multi-provider AI usage tracking setup for macOS

If you build with AI tools all day on macOS, you've probably had this moment: deep in a flow state, refactoring something, and Claude suddenly tells you you're rate limited. Or Cursor quietly stops working. Or you check your OpenRouter dashboard a week later and realize your credits are gone. The problem isn't the limits. The problem is the lack of visibility. Each provider has its own dashboard, its own reset window, its own pricing model. Nobody actually opens 8 tabs to check them. Here's the setup I landed on. The goal One glance, full picture. I want to know — without leaving my editor — how close I am to a rate limit, how many credits I have left, and when things reset. Across every AI provider I use. The tool I built a small macOS menu bar app called TokenBar that polls each provider's usage endpoint and surfaces the numbers in the menu bar. It supports Claude, Cursor, Codex, Copilot, Gemini, OpenRouter, Augment, Amp, JetBrains AI, Warp, Kiro, and a bunch of others — about 20 pro