%20--%3e%3crect%20width='1024'%20height='1024'%20rx='224'%20fill='%2316181d'/%3e%3cpath%20d='M%20292.8%20731.2%20A%20310%20310%200%201%201%20731.2%20731.2'%20fill='none'%20stroke='%238a909b'%20stroke-width='92'%20stroke-linecap='round'/%3e%3cpolygon%20points='700.1,329.6%20539.8,540.7%20484.2,483.3'%20fill='%23388bfd'/%3e%3ccircle%20cx='512'%20cy='512'%20r='44'%20fill='%23f2f3f5'/%3e%3c/svg%3e){kind=small}
Token Gauge
AI coding agent usage, in your menu bar
Know your limits before you hit them.
A live menu-bar meter for Claude Code, Codex, Cursor, Copilot and Gemini — Claude Code's 5-hour window and weekly cap counting down in real time, plus live usage for the rest. Local-only. $7.99 once.
Claude Code stays in the free tier — no account needed to start.
- Local-only — code never leaves your machine
- $7.99 once, no subscription
- 30-day money-back
- Builds on GitHub
30 days to change your mind. Email us, we refund you — no questions. No forms to argue with, no hoops. At $7.99, the only thing you risk is the minute it takes to install.
One developer, one tool, a real support inbox.
Token Gauge is made by Kieran Hartley, a working developer who got tired of guessing how much of each agent was left. There's a real person and a real support email behind it — not a growth funnel. Start free with Claude Code, no account required.
- Made by one developer who got tired of guessing.
- A real support inbox: kieranhartley@pm.me.
- Every build is open on GitHub Releases — you can verify what ships.
- 30-day money-back — the only thing you risk is the install.
One meter for every agent you run.
Stop finding out at line 200.
Claude Code runs on a 5-hour rolling window and a weekly cap. Token Gauge reads both live and shows the percentage and the exact reset minute — so you know before you start whether you can finish the refactor, instead of hitting the wall mid-task.
- Live 5-hour rolling window, as a percentage and a countdown
- Weekly cap with the exact minute it resets
- The number updates while you work — no dashboard to open
Five agents. One meter. Stop tab-hopping.
Claude Code, Codex, Cursor, Copilot and Gemini — all in a single menu-bar popover. Set your primary agent and its meter sits in the menu bar at a glance. No browser tabs, no four separate dashboards, no doing the math in your head across providers.
- One popover shows every connected agent at once
- Pick your primary — its gauge lives in the menu bar
- Switch focus without leaving your editor
See which agent and model eats your budget.
Where a provider exposes it, Token Gauge shows the detail: Cursor spend, Codex usage from what the tool reports, and per-model token breakdowns when they're available. You see the shape of your usage, not just a single bar — so you know which agent and which model is eating the budget.
- Per-model token table where the provider exposes one
- Cursor tracked by spend; Codex from its reported usage
- Honest about what's measured — see the fidelity table below
Warned before the cutoff, not after.
Set a threshold and Token Gauge notifies you as you approach the limit, while you can still wrap up the task cleanly. The whole point is foresight: the warning arrives in time to matter, not as an after-the-fact apology.
- Threshold notifications fire before you hit the wall
- Per-agent, so the alert is about the agent you're using
- You set the threshold — it fires once per crossing, not on a loop
Native on your Mac. The same app, same license, on Windows.
A small native menu-bar app — no browser engine bundled in, so it stays light. One license covers Mac and Windows, both included; Mac is never the more expensive one. On Windows it lives in the system tray.
- Lightweight native menu-bar app (Windows: system tray)
- macOS 12+ on Apple Silicon (M-series)
- Windows 10/11: standard NSIS installer — unsigned for now (More info → Run anyway)
- One license, both platforms — Mac is never the more expensive one
Your code and prompts never leave your machine.
Token Gauge reads usage stats locally. Over the network it makes only two kinds of calls: each provider's own usage endpoint, and one license check with Lemon Squeezy. No telemetry, no analytics, no account — there's no place for your code or prompts to go, because it never collects them.
- Reads usage on your machine; no telemetry
- Sends only provider usage calls + one license check
- No account to create, nothing to sign up for to start
Exactly what's tracked, for each agent. No pretending.
Different providers expose different things. We'd rather tell you the truth than imply one perfect real-time number across the board. We label every cell so you always know how much to trust the number.
| Agent | What Token Gauge tracks | Fidelity | Tier |
|---|---|---|---|
| Claude Code | 5-hour window, weekly cap, live %, reset countdowns, notifications | First-class authoritative — read-only usage endpoint; does not consume quota | Free |
| Codex | 5-hour and weekly rate-limits, plus tokens from the local rollout logs and live endpoint | First-class | Lifetime |
| Cursor | Request usage and limits, spend-based, from the local session DB or a pasted cookie | First-class spend-based | Lifetime |
| Copilot | Real per-SKU consumption: requests and credits this cycle, gross dollar value, billable overage | Best-effort real spend, no allowance % — GitHub doesn't expose the included quota size | Lifetime |
| Gemini | Token counts from the local CLI logs | Limited tokens only, no usage % — Google exposes no readable per-account quota | Lifetime |
A usage percentage is true only for Claude Code, Codex and Cursor. Copilot shows real consumption and spend, not a percentage. Gemini shows token usage. We label every cell so you never mistake an estimate for a hard number.
Note: Claude Code's free claude.ai plan has no usage endpoint and reads blank — Pro, Max, Teams, Enterprise, Enterprise Console and the API Console are supported.
We out-clarify. We don't out-feature.
Facts you can verify, not a knock on anyone. The most polished trackers in this category do excellent work; several are Mac-only, some cover a single agent, and some charge a recurring subscription. I made different, deliberate choices.
| Attribute | Token Gauge | A typical category tracker |
|---|---|---|
| Agents covered | Five — Claude, Codex, Cursor, Copilot, Gemini | Often one, sometimes a few |
| Platforms | Mac and Windows | Frequently Mac-only |
| Pricing | One $7.99 lifetime price | Sometimes tiers, sometimes a subscription |
| Privacy | Local-only; code and prompts never leave the machine | Varies |
| Devices | 3 | Varies |
| Updates | Lifetime, included | Varies |
| Refund | 30-day money-back | Varies |
Where others are ahead — you should hear it from me.
The most mature Mac-only trackers have been at this longer. They offer live cross-device sync over iCloud; Token Gauge doesn't sync today (cross-machine setup is plain settings Export and Import), and its history horizon is about 31 days, not 90. If you need that maturity and sync, those tools are a fair buy.
Our verifiable edge is simple: five agents, local-only, one flat $7.99 price, and it runs on Windows too. That's a stance, not a dig. If a Mac-only single-agent tool is exactly what you want, several good ones exist — buy one of those instead.
“I built Token Gauge for one reason: I was tired of finding out I'd run out of an agent halfway through a task. It tells you the truth about what each provider exposes, it keeps your code on your machine, and it costs $7.99 once. That's the whole pitch.”
Kieran Hartley · maker of Token Gauge · kieranhartley@pm.me
New tool, no borrowed reviews. Here's what you can verify instead:
- 30-day money-back. Email and you're refunded — no forms, no hoops.
- Local-only. Two network calls only: provider usage + one license check.
- A real inbox. kieranhartley@pm.me reaches the person who builds it.
- Open builds. Every release is on GitHub — inspect what ships.
One price. No subscription.
One price, on purpose. No middle tier to puzzle over, no subscription quietly renewing, no upsell waiting after you buy.
Free
$0
- Claude Code: 5-hour window + weekly cap, live %, reset countdowns, notifications
- No account needed. Start in under a minute.
- This is your trial for everything else — see Claude Code work first, then decide.
Lifetime
$7.99 once
- Every agent: Claude Code, Codex, Cursor, Copilot, Gemini
- Every agent we add in the future, included
- Mac and Windows, both platforms
- 3 devices
- All updates, for the life of the product
- No subscription. Ever.
- 30-day money-back guarantee — email us and we refund you.
$7.99 is the price, not a launch promotion. If we ever raise it, we'll say so plainly instead of dressing it up as urgency. Checkout happens right here, without leaving the page.
Local-only, and you can check.
- Reads your agents' usage locally; no telemetry, no analytics, no account.
- Over the network it makes only two kinds of calls: each provider's own usage endpoint, and one license check with Lemon Squeezy.
- Your code and prompts never leave the machine — there's no place for that data to go because we never collect it.
- It gives you visibility and timing — it does not raise your limits, add quota, or lower your bill.
Compatibility, stated plainly. macOS 12 or later on Apple Silicon (M-series); the Mac build isn't notarized by Apple yet, so the first launch is a right-click then Open. The Windows installer isn't code-signed yet, so SmartScreen will warn — choose More info, then Run anyway; signing is in progress. One license covers both.
Built by Kieran Hartley · kieranhartley@pm.me · Read the full Privacy Policy
Straight answers, in one sentence each.
Yes, on Apple Silicon (M-series) running macOS 12 or later. The Mac build isn't notarized by Apple yet, so the first launch needs a right-click then Open — after that it opens normally.
Same app, same license; one purchase covers Mac and Windows together. The Windows installer isn't code-signed yet, so SmartScreen will warn on first run — choose More info, then Run anyway. Signing is in progress.
Claude Code, Codex and Cursor are first-class with a usage percentage; Copilot shows real consumption and spend but no allowance % (GitHub doesn't expose the quota size); Gemini shows token counts only. The fidelity table labels each one so you never mistake an estimate for a hard number.
No — and we won't pretend it does. Token Gauge gives you visibility and timing so the wall is never a surprise; it does not raise your limits, add quota, or change what you pay your providers.
Free is real here — Claude Code stays in the free tier — but $7.99 unlocks the other four agents in one meter, the primary-agent gauge in your menu bar, notifications across every agent, and the same license on Windows.
No. Token Gauge reads usage locally and over the network makes only each provider's own usage calls plus a single license check; your code and prompts never leave your machine.
Three devices on one $7.99 license, lifetime updates included, no subscription — you buy it once and own it.
Email us within 30 days and we refund you — no questions, no hoops.
Install it, open the menu-bar app, and sign in to the agents you already use — there's no Token Gauge account to create, and Claude Code works the moment you launch.
Know your limits before you hit them.
Start free with Claude Code, or unlock all five agents for $7.99 once — Mac and Windows, 3 devices, lifetime updates, 30-day money-back.
Lost your license key? Recover it here. Every build is also on GitHub Releases.