Mac extension (panel/pill/nub) to show LLM usage (github.com)

A new Mac extension called UsageNotch has been introduced, designed to monitor and display users' interactions with AI models like Claude Code and Codex directly on their screen. The extension attaches to the edge of a Mac's display, providing a visual meter that tracks how much of the rate-limit quota for each AI service has been used. Users can hover over the meter to expand it into a more detailed panel showing usage metrics, including a 5-hour ring, a weekly meter, and estimated spending. This tool enables quick access to essential usage data without needing to navigate away from current tasks. Significantly, UsageNotch enhances productivity for developers working with AI models by making usage monitoring seamless and easily accessible. It operates locally, ensuring that sensitive data is not transmitted over the network—there are no API keys or external calls involved. The app efficiently reads from local files generated by the CLIs of the respective AI providers, ensuring real-time updates and community-oriented privacy. Key technical features include the ability to customize the pill’s position on the screen, cache data to speed up loading times, and optional account checks for real-time accuracy in Claude Code usage. Overall, this extension could be a game-changer for developers looking to optimize their engagement with AI services while keeping an eye on their resource allocation.