If you've ever gotten blindsided by a Claude Code rate limit mid-session or discovered you were burning $60/month on Opus for tasks Sonnet handles just fine, you're not alone. Tokens 4 Breakfast (T4B) is a native macOS menu bar app that aggregates your entire AI stack into one calm popover — tracking spend, token usage, and rate-limit pressure across eight providers without requiring a cloud backend or even an account.

Built Local-First for Developers Who Actually Care About Privacy

The app stores everything in a local SQLite database on your Mac. No telemetry, no account creation, no data leaving your machine. When T4B reads Claude Code's session files from ~/.claude/projects/, it extracts only the usage metadata object — token counts, model name, timestamp, cost — and never touches your actual prompts or conversation content. The same applies to every provider: official usage/billing API endpoints return aggregate data only. There's literally no technical path in the app to access your chats.

Supported Providers and Zero-Config Onboarding

T4B supports Claude Code, Claude Web (claude.ai subscription tracking), OpenAI, GitHub Copilot, Cursor, OpenRouter, DeepSeek, and Mistral — eight providers total. Here's where it gets clever: Claude Code connects automatically by reading local session JSONL files with zero configuration. GitHub Copilot pulls from your existing CLI credentials. The free plan gives you one provider; Pro ($7.99 one-time) unlocks all eight. Most users are fully set up in under two minutes.

Rate-Limit Forecasting and Session Budgets That Actually Work

The velocity-based rate-limit forecasting is the killer feature for heavy Claude Code users. T4B tracks your usage every 5 minutes and projects when you'll hit your 5-hour or weekly limit — with hours to spare, not after you're already blocked. The ambient menu bar cup shifts from calm to amber to red as you approach limits. Focus Mode takes this further: set a $1, $2, or $5 budget for the next hour of work, and T4B alerts you at 80% consumption in real-time across all connected providers. This is specifically designed to stop runaway agent loops while the session is still open — something no dashboard can do once you've switched tabs.

Usage Insights: See Where Every Token Goes

New in version 1.6, the Pro-tier Usage Insights feature breaks down spending by project and model with cache-reuse and Opus-share analysis. It calculates locally and never reads your prompts. The stretch-your-limit guidance tells you exactly what's eating your weekly quota before it becomes another quiet invoice.

Pricing: One-Time $7.99 vs. Free Tier

The free plan is genuinely useful: one provider of your choice, full menu bar dashboard, 7-day spending chart, and subscription tracker for Claude Pro or ChatGPT Plus. The Pro upgrade at $7.99 adds all eight providers, velocity forecasting, per-provider budget alerts, CSV export with weekly summary sharing, priority email support, and lifetime updates. That's a one-time payment — no subscription model — with a 14-day refund guarantee.

Key Takeaways

  • Native Swift app for macOS 14.6 Sonoma+ (Apple Silicon & Intel supported)
  • Claude Code auto-detects from local files; GitHub Copilot pulls from CLI credentials
  • Rate-limit forecasting gives hours of warning before session cutoff
  • Focus Mode session budgets prevent runaway agent loops in real-time
  • $7.99 one-time Pro upgrade with 14-day refund guarantee

The Bottom Line

T4B solves a real problem that most developers don't realize they have until they're staring at an unexpected $47/month drift on their credit card statement. The local-first privacy model is refreshing in an era where every tool wants your data, and the zero-config Claude Code integration means if you're already using it, you literally have nothing to set up. Worth downloading just to see what you've been quietly spending.