Glint

Product Introduction

1. Definition: Glint is a lightweight macOS menu-bar application and system monitor tool designed specifically for developers. It acts as a real-time status dashboard for Claude Code AI coding sessions, visualizing key metrics directly in the user's peripheral vision on their Mac desktop.
2. Core Value Proposition: Glint solves the problem of context-switching during AI-assisted development by providing glanceable, local monitoring of Claude Code activity. It surfaces live status, token usage, session plans, and context window health, allowing developers to maintain focus on their code while keeping essential AI session data visible and accessible.

Main Features

1. Real-time Session Dashboard: Glint reads the local session logs from the ~/.claude/projects/ directory to display a live dashboard. It shows the current status (Idle, Thinking, Awaiting), the active tool in use (e.g., "Editing NotchView.swift"), per-turn output tokens, cost, elapsed time, and a visual context ring that mirrors Claude Code's auto-compact gauge. This provides complete AI coding session transparency without leaving the current application.
2. Three Display Surfaces (Notch Island, Floating Pill, Dock Side): Users can choose from three independent, toggleable display modes: a Dynamic Island-style overlay that hugs the macOS camera notch; a draggable, always-on-top floating pill that can be positioned anywhere on screen; or a Dock-side panel that expands into unused screen space. Each surface shows the most active session and allows for quick expansion to view multiple sessions.
3. Privacy-First Local Processing: All data processing and rendering happen entirely on-device. Glint exclusively reads from the local .claude directory. It employs zero telemetry, requires no user accounts, and makes only one network call—to validate the license key with Polar—ensuring complete session data privacy and security.
4. Local Log Parsing & Session Aggregation: The application uses direct, read-only access to the ~/.claude file structure to parse and aggregate data from multiple concurrent Claude Code sessions. It intelligently displays the most active session upfront while tucking others into a expandable list, showing each session's model (e.g., claude-opus-4) and reasoning effort level.

Problems Solved

1. Pain Point: Developers using AI coding assistants like Claude Code face context-switching overhead and lose visibility into the AI's operational status when the terminal or IDE is not in focus. This disrupts workflow and makes it difficult to track token consumption, cost, and the progress of complex, multi-step tasks.
2. Target Audience: The primary users are Claude Code developers, AI-powered software engineers, and technical professionals who use AI for code generation, refactoring, debugging, and exploration. It is especially valuable for those managing multiple AI sessions simultaneously or working on lengthy, token-intensive projects.
3. Use Cases: Essential for monitoring long-running code generation or refactoring tasks, coordinating multiple sub-agent sessions for complex projects, tracking token budgets and costs in real-time, and receiving subtle notifications when an AI session requires user input, all without interrupting the primary coding environment.

Unique Advantages

1. Differentiation: Unlike generic system monitors or terminal-based output, Glint provides a purpose-built, persistent visual interface tailored for AI coding sessions. It moves status information from a buried terminal line to the most accessible parts of the macOS GUI (menu bar, notch area), offering a significant improvement in workflow ergonomics over standard command-line interfaces.
2. Key Innovation: The core innovation is its fully local, privacy-centric architecture combined with a native macOS-native UI design. By processing ~/.claude logs on-device and offering interface elements like the Notch Island, it delivers a seamless, secure, and non-intrusive monitoring tool that feels like a natural part of the operating system, unlike cloud-based or web-wrapped alternatives.

Frequently Asked Questions (FAQ)

1. How does Glint protect my privacy and session data? Glint is designed with a strict privacy-first approach. It operates entirely on your local machine, reading session data exclusively from the ~/.claude directory. It sends no telemetry, requires no cloud account, and the only external network call is a one-time license key validation with Polar. Your code and AI interaction data never leave your Mac.
2. What AI coding tools does Glint support besides Claude Code? Currently, Glint is built specifically to monitor Claude Code sessions by parsing its local log files. Support for other AI coding providers and tools, such as OpenAI Codex, Amazon Kiro, and others, is listed on the developer's roadmap for future releases.
3. What are the system requirements and how do I install it? Glint requires macOS 14 Sonoma or later to support modern menu bar and notch features. It is distributed as a standard .dmg disk image. After dragging the app to your Applications folder, you may need to clear the macOS quarantine flag via a terminal command or through System Settings > Privacy & Security to allow the unsigned app to run.
4. What does the lifetime license include and how much does it cost? The Glint Lifetime License is a one-time purchase of $7 (discounted from $15), which grants you perpetual access to the software and all future updates. The license covers activation on up to 2 Macs. Payment and key delivery are handled through the Polar platform, with activation occurring directly within the Glint application.
5. How is Glint different from just watching the Claude Code terminal output? While the terminal shows output, Glint provides a persistent, glanceable dashboard that remains visible across all applications. It aggregates key metrics (token spend, cost, time, context usage) into visual elements like rings and progress bars, displays multiple sessions simultaneously, and offers non-intrusive notifications—capabilities a static terminal window cannot match without significant scripting.