Codenotch

Codenotch is a desktop monitor for people who use AI coding assistants and local LLMs. A small display at a screen edge keeps usage readings and session states visible while you work in other apps.

• Assistant usage: Check limits and reset times for Claude Code, Cursor and Codex, plus account readings from DeepSeek, GLM, MiniMax and Grok; Antigravity may show only a request count when quota readings aren't available.
• Local model activity: See loaded Ollama and LM Studio models and their context use; Ollama shows memory information, while optional response capture adds thinking activity and generation speed, and LM Studio shows prompt processing, generation, queues and daily token counts.
• Session alerts: Get a visual and audible alert when an agent finishes or needs input, then bring its application to the foreground by clicking the display.
• Reading quality: Distinguish provider-reported figures from derived or manual readings and see a visible status when a lookup fails.
• Phone display: View the Mac's usage percentages, reset times and session states on iOS or Android over the same Wi-Fi network; the phone doesn't receive tokens, credentials or raw API responses.

Codenotch runs on macOS 15 or later and has a Windows port. It's open source under the MIT license. Cloud usage checks can contact provider services, so those readings need internet access. Many integrations reuse an existing sign-in; some require a separate login or API key. Session alerts select the application, not a particular terminal tab.