TrackNotch

What is TrackNotch?

TrackNotch is a native macOS app that shows real-time LLM usage in your Mac’s notch or, on Macs without a notch, in the top menu bar. It is designed to monitor usage across multiple providers, including Claude, OpenAI, Cursor, Codex, Anthropic API, and Google Gemini, and present that information in a compact interface while keeping the data local to the machine.

The app reads provider usage from local files and APIs rather than routing traffic through a proxy. It also supports local tracking for Claude Code session context usage and can surface 5h/7d rate-limit data when an Anthropic OAuth token is provided. TrackNotch stores API keys in the macOS Keychain and is documented as having no cookies, telemetry, or third-party analytics.

Key Features

• Notch or menu bar display — shows LLM usage in the notch on notched Macs, with a menu bar fallback for other Macs.
• Multi-provider tracking — monitors usage for Claude Code, OpenAI API, Cursor, Codex, Anthropic API, and Google Gemini in one interface.
• Local-first data access — reads from provider files and APIs on the machine, so tracked data stays local unless the app needs to contact a configured provider API.
• Claude Code context arc — displays a live visual arc for the current Claude Code session’s context window usage.
• Budget and spend tracking — supports monthly budget tracking for OpenAI and Anthropic when admin-level keys are used.
• Keychain-backed secrets storage — stores API keys in macOS Keychain instead of plaintext files.
• Rate-limit header support — can use Anthropic OAuth token data to show 5h/7d Claude rate-limit information.

How to Use TrackNotch

Install the DMG release, move TrackNotch.app into Applications, and open it. Because the build is unsigned, the first launch may require opening it through Finder’s context menu or removing quarantine attributes in Terminal.

After launch, open the pill in the notch or menu bar and go to Settings. Paste in the API keys or OAuth token for the providers you want to track, then let the app refresh automatically. For Claude Code context tracking, the app can read local session files without an API key; rate-limit tracking requires an OAuth token.

Use Cases

• A developer checking Claude Code context usage — see how full the active session context window is before continuing a long coding task.
• A user monitoring multiple AI services — keep an eye on Claude, OpenAI, Cursor, Codex, or Gemini usage from a single status element instead of checking each service separately.
• Someone tracking monthly API spend — use admin-level keys to watch OpenAI or Anthropic spending against a set budget.
• A macOS user who wants low-friction status visibility — display usage in the notch or menu bar without opening a separate dashboard.
• A privacy-conscious user — rely on local reads and Keychain storage instead of a proxy-based monitoring setup.

FAQ

Does TrackNotch work only on notched Macs?
No. It is designed for the notch on supported Macs, but it also includes a menu bar fallback for Macs without a notch.

Does it require sending data to a third-party service?
According to the repository, TrackNotch keeps data local, reads from provider files and APIs, and does not use telemetry or third-party analytics.

Do all providers need API keys?
No. Claude Code and Cursor can be tracked through local monitoring, while other providers use API keys or OAuth tokens depending on the feature.

Can it track Claude Code rate limits?
Yes, but the repository indicates that 5h/7d rate-limit data requires an Anthropic OAuth token.

Is the app sandboxed?
No. The repository states that it is intentionally unsandboxed so it can read provider data from locations outside the App Sandbox, such as local configuration directories.

Alternatives

• Provider dashboards — web-based usage pages from OpenAI, Anthropic, or other services can show account or API usage, but they are separate per provider and do not combine everything into one macOS status view.
• Manual monitoring workflows — users can check local files, provider consoles, or billing pages directly, which avoids installing another app but does not provide a live notch or menu bar display.
• Custom scripts or terminal tools — a script can read provider data and summarize usage, but it requires setup and does not provide the polished, always-visible macOS interface that TrackNotch offers.
• General macOS menu bar utilities — other status bar apps can surface system information, but they may not be focused on LLM usage tracking or provider-specific context and budget data.