What is running?
See observed activity and estimated output speed, with units beside each number. Open the live modules to inspect the tool and its tasks. When no tool is working, the speed display stays quiet.
A CLEARER VIEW OF YOUR AI TOOLS
Your AI usage.
In clear view.
Activity, output speed, and remaining allowance, together in your menu bar. Open a module for detail. Your usage history stays on your Mac.
Signed and notarized for macOS. Open the .pkg to install; no build tools or reboot required.
Apple silicon · macOS 14 or later · MIT licensed
Local by default. Ready at a glance.
NATIVE APP PREVIEW
One place for what’s happening now.
Menu bar
Estimated output speed while tools work. Dated allowance readings alongside it.
Live modules
A consistent place for each tool’s activity, output speed and allowance.
Quiet when nothing is running
Remaining stays named on the menu bar. Unused zeros stay off.
Closer look
Expand a module for its reading, reset time and activity details.
Each module opens in place. Reports brings together History, Cost and Allowances; Settings stays close at hand. Missing readings remain explicit. Optional allowance warnings appear on the affected tool.
A SMALL APP FOR YOUR WORKDAY
Know what’s happening.
Know what’s happening.
Keep what you’ve learned.
What allowance remains?
Live modules retain supported account allowances through idle periods. Read age and missing values stay explicit. Optional warnings use account-bound allowance observations; forecasts remain separate from output speed.
Where did the usage go?
Explore recorded usage by tool, task, model, project or day. Compare supported usage against API prices, or export your records to CSV. Cost estimates are labeled separately from account allowance.
COST & CONTEXT
Put cost in context.
Put cost in context.
Compare what produced it.
Output includes reasoning and does not measure answer quality. API estimates, local tokens, account totals and allowance changes remain separate measurements.
Compare cost with output.
The Cost view compares full estimated request cost per million output tokens, by day and model. Input and cache costs stay included. Only records with a supported price enter both sides of the calculation, with output coverage shown alongside it.
Separate rate changes from usage changes.
Inspect each model’s shipped API price history: effective dates, input/cache/output prices, changes and published sources. Historical estimates use those dated rules; reference estimates hold the card fixed. The last verification date stays visible, and unknown intervals remain unpriced.
Observe allowance alongside tokens.
Compare observed Codex allowance changes with local input, output and cache use over the same account/window intervals. Resets, gaps and ambiguous attribution stay visible. Model mix provides context, but does not reveal a model-specific subscription multiplier or identify another host.
Your history stays. New records join it.
Use your tools as usual.
Token Bar reads available local usage from Codex, Claude Code, Grok, and OpenCode. Coverage depends on what each tool records.
Keep what you’ve collected.
Saved measurements survive app launches. Repeated records count once, and incremental reads pick up changes without starting your history over.
Open the view you need.
History and Cost reuse prepared reports. Allowances keeps provider readings separate. Switching reports retains your choices while new records arrive in the background.
Account quota follows a separate path: supported account observations → remaining allowance and reset times. Missing, stale, and unpriced data stay labeled.
YOUR MAC. YOUR RECORDS.
Local by default.
Local by default.
Inspectable by anyone.
Usage history stays on your Mac. Reports are drafted locally. You choose when to send the draft to GitHub for review.
Read the privacy detailsA few things to know.
Does this measure billing or exact subscription costs?
No. Processed-token totals are local usage records, not a bill. Quota is read separately from the connected account. Speed and exhaustion time are estimates, with freshness and coverage shown in the app.
What do I need to run it?
An Apple silicon Mac running macOS 14 or later. Token Bar reads local usage from Codex, the local Grok client, Claude Code, and OpenCode. Install it locally and use a supported ChatGPT sign-in for Codex account quotas. API-key-only and unsupported account responses may have no quota reading. Local logs and account APIs can change between releases.
Do all sources provide the same readings?
No. All four sources provide token usage and history. Account attribution and quota are available for supported Codex accounts; Claude quota uses account-bound local data that the installed Claude Code refreshes every 15 minutes, including its Fable weekly limit. Output speed is available for Codex, partial for Grok, estimated for Claude Code, and unavailable for OpenCode. Codex cost estimates use dated API rates; OpenCode token history is separate from its provider-reported billing. Claude Code and Grok usage remain unpriced. Missing readings stay unavailable.
Does it see work on my other computers?
Token Bar reads the usage records available on this Mac. It does not automatically collect logs from other hosts or cloud tasks, and saved records do not identify the originating host. Supported Codex account totals can provide account-wide context, but cannot locate that usage. Missing remote records stay unknown.
Can I customize what appears?
Yes. Choose the icon, activity counts, output rate, amount left, projected zero, speed dial, and optional Fable quota with its time left at recent pace. Reorder them, choose compact labels, and set rate units. Preferences save automatically.
Is this affiliated with an AI provider?
No. Token Bar is an independent open-source utility. It is not affiliated with or endorsed by any AI provider.
Install Token Bar.
Download the signed, notarized .pkg and follow the macOS installer. Token Bar goes into Applications, ready to open on an Apple silicon Mac with macOS 14 or later.
Your saved history and preferences stay in place. The app source, synthetic tests, build scripts and this website are also available on GitHub.
Keep your work in view.
Small footprint. Useful signals. Source you can read.
Explore on GitHub