Pulse

• Seventy-seven services at a glance — Claude Code, Codex, Cursor, GitHub Copilot, Antigravity, Kiro, Grok, DeepSeek, Kimi Code and many more, each as its own ring.
• No Pulse account, no Pulse server — it uses the logins you already have, and nothing is sent back.
• Real numbers only — every usage percentage is the one the service itself reported. The few estimates are marked as such, and where a service reports nothing, Pulse says so instead of guessing.
• Native and quiet — Swift and SwiftUI, Liquid Glass on macOS 26. Dock it left, right, along the top or along the bottom beside the Dock, let it fold to a sliver when idle, or keep it in the menu bar.

Recently added: usage in the menu bar with a dashboard for each account, starting usage windows after a reset, and animated bot marks. What's new.

Key Features

At-a-Glance Status Rings

• Usage-Aware Colors: Dynamic color gradients shift from green to amber, red, and deep red when exhausted — or set custom accent colors per account.
• Active Turn Indicator: A subtle revolving dot indicates whether an agent is actively generating responses in real-time (Claude Code, Codex, Kiro, Zhipu and z.ai).
• Window Clock Arc: An optional secondary outer arc can show how much of the current rate-limit window has elapsed or remains.
• Countdown Mode: Toggle between showing spent quota (75% used) or remaining balance (25% left).

Hover Details & Smart Forecasting

• Complete Limit Breakdown: Hover over any ring to reveal a detailed card showing every reported quota pool, reset countdowns, and current window status.
• Detailed Card (Optional, per account): Switch it on for the accounts you watch closely. The card adds the plan, when the figures were read, an estimate of what each limit is worth wherever this Mac's records can price it, and the account's recent usage — today, 7 days and 31 days of tokens, a 31-day chart, the model doing most of the work, the cache hit rate where the records sort every token by kind, and how long the prompt cache has left — for Claude Code from the tier each reply records, for Codex as the at-least-30-minutes OpenAI guarantees on GPT-5.6 and later: the card names the conversation about to lapse, and Settings lists every one — for every service Pulse has a history for: z.ai and Zhipu from their own account statistics, and Claude Code, Codex, Kimi Code, Grok, OpenCode, Cursor, Devin, Antigravity, Command Code and Copilot from this Mac's records with Token spend on.
• Burn-Rate Forecast (Optional): When you turn it on, projects whether your current pace will outlast the quota window and shows an estimated time-to-exhaustion (ETA) when risk is detected. Off by default.
• Pin Primary Window: Pin whichever limit matters most to the ring, or let Pulse automatically track the one closest to exhaustion.

Native, Fluid & Non-Intrusive

• Flexible Edge Docking: Dock to the left, right, top (above the menu bar) or bottom of your screen (beside the Dock), or float it anywhere, standing upright or lying across.
• Multi-Monitor Native: Drag Pulse to any secondary display; it remembers screen placement and gracefully returns if disconnected. Turn on Follow the active display and the single rail moves itself to whichever screen your pointer is on.
• Auto-Collapse: Automatically folds into a razor-thin sliver when idle to eliminate distraction, glowing red only when quota runs critically low.
• Opt-In Notifications: You choose which ones to switch on. Get told when a limit passes 75/80/90/95%, when the provider says it is spent, when a window you were warned about comes back, and when several checks in a row fail so the panel is quietly showing older figures, and — for the services that sell prepaid credit — when the balance falls under a figure you set, and when Codex's, Claude Code's or DeepSeek's own status page reports what the tool runs on as down (only for the ones you have switched on). Each thing is said once: a limit already past the line when you switch this on is mentioned straight away, and again when it resets or gets worse.
• Start Usage Windows After a Reset (Optional): Claude Code's and Codex's windows only start at your first message after a reset. Switch this on and Pulse sends a single "hi" through the provider's own command-line tool just after each reset, within the hours you choose, so the clock starts then rather than whenever you come back. Off by default, and it is not a feature of Anthropic or OpenAI: switching it on asks you to confirm that it may be treated as getting around usage limits.
• Spaces-Friendly: Keeps to the Space you are working in, leaving full-screen apps to themselves.
• macOS Aesthetic: Classic solid obsidian surface or native Liquid Glass on macOS 26+.
• Animated Marks (Optional): Replace a provider's logo with a small bot that reacts to what that account is doing — working, fetching, spent, or quiet. Off by default and switched on per account, with eight personalities, eighteen body shapes and a colour of your own if you want one.
• Rail Menu & Shortcuts: Right-click the floating rail — or the collapsed sliver; Control-click works too — for a menu with Settings and Quit. In Settings › General › Shortcuts you can optionally assign global shortcuts to Open settings and Show or hide the panel; both are unset until you set one.
• Menu Bar Usage (Optional): Show the fullest ring — or an account you pick — beside the menu bar icon as a figure, a small ring, or its five-hour and weekly limits side by side (5h/9% Wk/15%), red past the warning line. Turn on Usage panel in the menu and clicking it opens a dashboard: an overview of every account, and a tab per account with each limit, its reset, credits, estimated spend (with Token spend on) and a link to the provider's own usage page. Turn the floating panel off from the same menu if you only want the menu bar.
• Five Interface Languages: English, Simplified Chinese, Traditional Chinese, Japanese and Korean, with language-aware large-number units: K/M/B, 万/亿, 萬/億, 万/億 and 만/억 respectively.

Multi-Account & Local Ledger

• Multi-Account Support: Monitor multiple subscriptions for the same provider (Claude Code, Codex, Grok, Grok Bot) side-by-side with custom labels.
• Token Spend (Settings-only): Off by default. Enable it on the page to scan local records; switching it off stops the scan. Reads local logs, databases and exports from a catalogue of 54 client sources, including Gemini CLI, Cline, Roo Code, OpenClaw and GitHub Copilot. Cursor, Trae and other export sources need a prior export or capture. These are distinct from the rail's 77 quota providers; support and live-client validation vary by source. Sources and coverage.
• Clear Usage Estimates: Opens on the last 7 days and remembers your chosen span. Costs use published API prices, not subscription charges. Unknown prices stay unavailable, incomplete counts and coarse timing are labelled, and a source that reports no token counters says so.
• Model Details & Charts: Open a model for input/output/cache counts and estimated costs, daily and hourly charts where the records support them, contributions by agent, and sortable, paged detail tables. Point at a chart to read the date or hour and its token count. Unavailable daily or hourly detail is shown as unavailable, not zero.
• Monthly and Yearly Recap: A set of cards for any month or year, built from the same local records as Token spend and made to be shared — tokens and estimated cost against the month before, every day on a calendar, the hour you work most, the models, agents and projects behind it, your stre