AgentNotch is a sleek macOS menu bar application designed for developers who use AI coding assistants like Claude Code and OpenAI Codex. It provides real-time telemetry and visibility into AI assistant activities directly from the Mac's notch area, allowing users to monitor tool usage, token consumption, and costs without interrupting their workflow. The app is specifically built for macOS users who want to maintain awareness of their AI coding processes while focusing on other tasks.
As AI coding assistants become more integrated into development workflows, developers often lack visibility into what these tools are doing in real-time. This creates uncertainty about whether the AI is still processing, what files it's accessing, what commands it's executing, and how much it's costing. AgentNotch solves this problem by providing a persistent, unobtrusive monitoring interface that shows exactly what Claude Code and OpenAI Codex are doing at any given moment, addressing the need for transparency and cost control in AI-assisted development.
One key feature is real-time tool tracking, which displays every tool call as it happens including file reads, code edits, shell commands, and other actions. This allows developers to know exactly what their AI assistant is doing at any moment without switching contexts. The app monitors tool usage continuously and presents this information in an accessible format that expands on hover to show detailed call histories.
Another important capability is token and cost monitoring, which tracks token usage for both input and output while providing estimated costs in real-time. This feature helps developers avoid unexpected API bills by giving them immediate visibility into consumption patterns. The monitoring includes breakdowns that help users understand the cost implications of their AI coding activities as they happen.
The app features source-aware design with color-coded indicators: orange for Claude Code, blue for Codex, and light blue for unknown sources. This visual distinction lets users know which AI is active at a glance without needing to examine detailed logs. The color system provides immediate identification of which assistant is currently processing tasks.
AgentNotch includes completion detection that notifies users when their AI assistant finishes a task, eliminating uncertainty about whether the AI is still thinking. This feature provides clear signals about task completion status, so developers don't need to constantly check their editors or terminals to see if AI processes have concluded.
admin
The app is lightweight and native, living in the Mac's notch to use zero screen real estate while providing fast, efficient, battery-friendly performance. It expands on hover to show details and collapses to a minimal indicator when not needed. This design ensures the monitoring doesn't interfere with other applications or workflows.
AgentNotch works by receiving telemetry data from AI coding assistants through OTLP/HTTP protocols on port 4318. It listens for OTLP logs and metrics from configured tools, decoding the information to display real-time activities. The app integrates with Claude Code through environment variable configuration and with Codex through configuration file settings, creating a local monitoring system that processes telemetry without external dependencies.
Users benefit from increased transparency into AI coding processes, better cost control through real-time token monitoring, and reduced context switching since they can monitor AI activities without leaving their current tasks. The app provides peace of mind by showing exactly what AI assistants are doing and when they complete tasks, while also helping optimize API usage and expenses through continuous visibility.
Common use cases include monitoring Claude Code sessions during complex refactoring tasks, tracking Codex usage while generating boilerplate code, observing file access patterns during AI-assisted debugging, and monitoring token consumption during large code generation sessions. Developers can use it to ensure AI tools aren't accessing sensitive files, to understand cost patterns across different types of coding tasks, and to verify that AI processes have completed before moving to the next task.
The target audience includes developers using Claude Code or OpenAI Codex on macOS, particularly those working with multiple AI assistants who need to distinguish between them. The app integrates with existing AI tool configurations through simple setup steps and requires macOS 14.0 or later, with optimal experience on Macs with notches (MacBook Pro 14\" or 16\" 2021+) though it falls back to menu bar on other Macs. It runs 100% locally with no data sent externally.
AgentNotch provides essential visibility into AI coding assistant activities through an unobtrusive interface that helps developers monitor, understand, and control their AI-assisted workflows while maintaining focus on their primary development tasks.
AgentNotch targets developers using AI coding assistants like Claude Code and OpenAI Codex on macOS systems. The primary users are software engineers, programmers, and technical professionals who integrate AI tools into their development workflows and need visibility into what these assistants are doing. It's particularly valuable for developers working with multiple AI assistants, those concerned about API costs, and professionals who want to monitor AI activities without constant context switching. The app requires macOS 14.0 or later and works optimally on MacBook Pro models with notches (2021+ 14" or 16" models) though it functions on any compatible Mac.