AI Token Usage Tracker

The Intraday tab shows today's API costs per key with approximately 1 hour delay from Anthropic's Usage API. Each API key is displayed as a card with per-model breakdown, color-coded cost distribution bars (Input, Cache Creation, Cache Read, Output), and estimated totals.

A dedicated refresh button lets you pull the latest intraday data independently from the rest of the app. If you've configured a spend limit, a combined progress bar shows your month-to-date cost plus today's estimate against your budget.

The By Month tab shows paginated month cards (2 per page) with costs from Anthropic's Cost API — finalized dollar amounts with approximately 2 days of reporting lag. Each month card displays the total cost, a color-coded cost distribution bar, and a 2×2 grid showing per-category costs (Input, Cache Creation, Cache Read, Output) with percentages.

Pro users can expand each month to see a per-API-key breakdown — isolating exactly how much each project, bot, or team member spent. An all-time grand total appears at the bottom across all months.

The By Model tab lists every Claude model you've used — Claude Opus, Sonnet, and Haiku in all versions — sorted by version (newest first) and tier. Each model card shows total cost, a color-coded distribution bar, and a stat grid with the 4 cost categories (only those with actual usage).

If you've configured rate limits, each model also displays inline rate limit pills showing requests per minute, input tokens per minute, and output tokens per minute. The app uses dynamic model discovery via the Anthropic Models API, so new Claude models are supported automatically.

In Settings, you can select which API keys to include in cost calculations. This lets you isolate costs by project (e.g., only show costs for your production bot) or compare spending across different API keys used by different team members or services.

The filter applies across all tabs — Monthly, Model, Workspace, and Intraday. Free tier supports 1 API key; Pro unlocks unlimited keys.

The By Workspace tab shows per-workspace cost breakdowns from Anthropic's Cost API. Each workspace is displayed as a card with its name and total cost, making it easy to compare spending across Production, Development, Staging, or any other workspace in your organization.

An "All Workspaces" total appears at the bottom for a quick organization-wide overview.

Enter a regular Claude API key in Settings to probe your current rate limits. The app sends a minimal 1-token request per model (costing approximately $0.0001) to read the rate limit headers returned by Anthropic. Results show requests per minute, input tokens per minute, and output tokens per minute for each model.

Rate limits are displayed inline on model cards and in a dedicated table in Settings. Results are cached across sessions, and the feature supports all models discovered via the Anthropic Models API.

Set a monthly spend limit (in USD) that mirrors your Anthropic Console setting. A visual budget progress bar shows your current spend as a percentage — green below 75%, orange at 75–90%, and red above 90%.

Enable spend threshold notifications to receive a native macOS notification when your costs reach a configurable USD amount. Notifications fire once per calendar month (per threshold). You can test the notification from Settings to make sure it's working.

Switch between 12 currencies with a single click via the flag-based dropdown in the header: USD, EUR, GBP, JPY, CNY, CHF, CAD, AUD, KRW, INR, BRL, and SEK. Exchange rates are fetched in real-time from an open exchange rate API and cached across sessions.

All views, CSV exports, and the JSON API automatically use the selected currency. You can also configure a custom exchange rate endpoint for enterprise or air-gapped environments.

Export cost data from any tab with one click. Each tab produces a tailored CSV format: Monthly exports include month, token type, tokens, and cost (plus per-key breakdown for Pro). Model exports list model, token type, tokens, and cost. Intraday exports include API key, model, token type, tokens, and estimated cost.

All amounts are exported in your selected currency with proper decimal formatting. A grand total row is appended at the bottom for easy accounting.

Enable the built-in localhost-only HTTP server (default port 9485) to expose your usage data as JSON. The API provides a full snapshot including monthly costs, per-model breakdowns, workspace totals, intraday data, and spend limit status — all in your selected currency.

The server listens exclusively on 127.0.0.1 (no external network exposure) and supports CORS for browser-based dashboards. Endpoints: GET / or GET /usage for full data, GET /health for status checks. Test with: curl http://localhost:9485

By default, the app appears as a menu bar popover that closes when you click elsewhere. Pin the window to switch to detached mode — a floating window that stays visible while you code, switch apps, or work in the terminal.

Toggle between popover and floating window with the pin button in the header. Your current cost is also optionally displayed directly in the macOS menu bar text.

Your Admin API key is stored in the macOS Keychain — never in plaintext or UserDefaults. The app communicates only with Anthropic's official API endpoints and an exchange rate service. There is no analytics, no telemetry, no crash reporting, and zero data sent to any third party.

The entire app is built on Apple system frameworks only — zero third-party Swift packages or CocoaPods. The JSON API server is bound to localhost only, preventing any external network access. API keys are masked by default in the Settings panel with a show/hide toggle.