82%5-hour
5-hour 7-day
Resets in 2h 40m
✦DEVS! Reduce your Fable 5 costs Track ![]()
AIlimits
before they stop you.
Usage, resets, costs, and alerts across your AI stack.
Mac app and web dashboard. Windows coming this summer.
★★★★★
Our systems are trusted by 25,000+ AI heavy users.One tiny island. Your whole stack.
It lives in your menu bar, expands into every source you track, Fable 5 included, and turns amber then red before a limit can stop you.
The tokenkarma Mac app lives in the menu bar as a compact island. It expands into a live panel showing every AI source you track: session, daily and weekly windows per provider, including Anthropic's new Fable 5 tier with its own weekly window, real API spend collected server-side, alerts at 75, 90 and 100 percent, and one click to the full dashboard.
The invisible AI tax stops here.
$66
average monthly AI spend
per subscriber
per subscriber
4
different AI tools
per subscriber
per subscriber
25%
spend over $100
every month
every month
Two subscriptions here, an API key there, a tool you forgot you pay for. tokenkarma adds it up: subscriptions, API spend, every provider, one number.
Source: Bango, Nov 2025 (US consumer survey).
Cost over time it adds up to$127.43/mo
one tool at a time a subscription here, an API key there
Mac. Web. One account.
The island in your menu bar, the dashboard in your browser. All synced server-side: connect once, see the same truth everywhere.
One dashboard to rule them all.
Real cards from the composable dashboard. Pick the ones you want, build your own view.
Near limit 2 alerts
7590100
Claude at 90%. Plan around it, your window resets at 6:10 PM.
Peak hours 5-11 AM PT
Year of activity 61 sessions / wk
Projected June spend ~$62
July 7: Fable 5 leaves the included plans.
Cap this session's extra usage. One command.
Fable 5 goes 100% extra usage on July 7. Anthropic caps your spending once a month. tokenkarma caps it per session, per day, per week, and blocks the prompt before it spends.
tokenkarma, your token's guardian.
> /tokenkarma 20
Session capped at $20.00 of extra usage (counted from now)
┌─ 🛡 tokenkarma ───────────────────────────────
│ 🔴 EXTRA USAGE — limits armed
│ /tokenkarma session|day|week <usd>
│
│ Session $0.38 / $20.00
│ █░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░ 2%
│ Day no limit
│ Week no limit
│ Extra usage this month 178.26 / 500.00 EUR
│ ███████████░░░░░░░░░░░░░░░░░░░░░░░░░░░░░ 36%
│
│ ⛔ 1 block today
└─ ⚡ /tokenkarma 20 · /tokenkarma menu ─────────What users say
25,000+ people track their AI limits with our engine
Now, I don't need to worry about running out of limits all of a sudden. I can track the limits and plan my prompts accordingly.
Ishan Mendiratta 
Finally! Huge time saver. Thanks guys!!
Christian Harten Perfect UI: the most important information is available at a glance... more detail when you click, not overwhelmingly so. Well done.
Richard Bartholomew Perfect tool, works in 1 click
Lucas Israel the best tracker .. but i wish that i can understand what karma points do :D
Fathy Yakot Simple (but feature-rich), lightweight, and easy to use; love it!
Ishraq Very easy to use... with real time updates, warnings and incredible UI.
Felix Kabira Kagwi very useful for pro devs
Arnaud Tartulier You keep the keys.
AES-256-GCM encrypted at rest, unique IV per secret
Masked display last 4 characters only
One-click revoke hard delete, effective immediately
Sessions visible see them, kill them yourself
GDPR self-service export or erase, no email needed
No passwords Google, GitHub or 6-digit code
The subscription that pays you back.
tokenkarma finds the waste in your AI stack: idle subscriptions, wrong plans, quota you pay for and never use. Every dollar you put in comes back off your AI bill. Early Access pricing applies automatically to your first year on an annual plan, while launch spots remain.
2 months free on annual billing
Prices in EUR. Cancel anytime. Card required for the trial. Early Access pricing applies automatically to your first year on an annual plan, while launch spots remain. Renews at the standard annual price. Monthly plans are available at the standard price.
Managing a team? Join the Team waitlist.
Tell us what you want to do. We give you the keys.
The AI Cost Strategist is what we are building next: describe the task, and tokenkarma recommends the right model, the right workflow and the right tool for it, grounded in what you actually use and spend. It is the most requested feature from our early users.
1 · You describe
Plan two weeks of social content for our launch.
2 · Grounded in your usage
3 · We recommend 
The right modelmatched to the task, not to habit
The right workflowstep by step, with the prompt to start
The right toolfrom the stack you already pay for
Frequently asked questions
- What does tokenkarma track?
- tokenkarma tracks two kinds of AI usage in one place. First, subscription limits: the caps built into plans from Claude, ChatGPT, Gemini, Grok, Perplexity and Cursor, including 5-hour sessions, daily and weekly windows and per-model quotas, collected live with the reset time for each window. Second, API spend: for key-based providers, tokenkarma reads the cost reports your providers already publish and totals what your keys actually burn. Beyond the six first-class providers, over 100 AI tools are catalogued, so a subscription you forgot about still shows up in your stack. Everything lands in one dashboard: live usage, predicted time to cap, reset schedule and monthly cost, with alerts before a limit stops you.
- Do I need an API key?
- No, not for subscription tracking. tokenkarma reads the usage caps of the plans you already pay for: sessions, daily and weekly windows and per-model limits are collected without any API key. An API key is only needed if you also want API spend tracking, where tokenkarma reads the cost and usage reports exposed by your providers to show what your keys actually cost. Any key you add is encrypted at rest with AES-256-GCM, displayed masked (last four characters only), used for one thing, reading usage and cost data, and revocable in one click. Revocation is a hard delete and takes effect immediately. The security page documents the full handling.
- Which platforms does tokenkarma run on?
- tokenkarma runs on two surfaces synced to one account. The native Mac menu bar app (macOS 13 and later) lives next to your clock as a compact island and expands into a live panel of every source you track. The web dashboard gives you the full picture in any browser: history, alerts, settings and your whole stack on one screen. Add a source once and it appears everywhere, with the same numbers and the same reset times. A Windows app is coming this summer; join the list on the Windows page to get one email the day it ships.
- Is my data safe?
- Yes, and you stay in control of every credential. Secrets are encrypted at rest with AES-256-GCM with a unique IV per secret, decrypted only transiently for collection runs, never written to logs, and always displayed masked. Revoking a credential is a hard delete, effective immediately. Device tokens are read-only and revocable one by one. Your active sessions are listed in your account and you can terminate any of them yourself. Under GDPR you can export everything tokenkarma holds about you and erase it, both self-service, no support email required. The infrastructure runs on Cloudflare in the EU region. The security page documents the architecture, and the privacy policy lists exactly what is collected.
- What does tokenkarma cost?
- tokenkarma is a paid subscription designed to pay for itself: it surfaces the waste in your AI bill, idle subscriptions, wrong plans and quota you pay for but never use, so the savings it finds are meant to exceed its price. Every paid plan starts with a 7-day free trial. Two paid tiers cover different stacks: Hobbyist for a couple of sources, and Expert for heavy users who want API spend tracking, full history, alerts across their whole stack and the Claude Code spending caps. Prices are served live from the pricing API and shown on the pricing page, including the current Early Access pricing for annual plans.
- How does tokenkarma save me money?
- Three mechanisms, and only measured savings are ever counted. Auto-switch runs simple background tasks in Claude Code on a cheaper Claude model automatically, then reprices the real tokens consumed and credits that measured amount to your total; in our internal model-mix benchmark, running routine work on cheaper tiers cut model costs by 47% to 86%. Cheaper-model tips suggest a cheaper Claude tier on prompts where it would do the job, in any language, and the saving is credited only when you follow the tip. Spending caps stop extra usage at the ceiling you set per session, day, week or month, where Claude itself only offers a monthly cap. If nothing was switched and no tip was followed, nothing is credited: the savings number is a measurement, not an estimate.
- Does tokenkarma switch models for me?
- Only if you turn auto-switch on, and only downward. By default tokenkarma is read-only: it shows your live usage and costs, predicts when you will hit a cap, and alerts you before that happens, without touching your requests. Auto-switch is an opt-in Expert plan feature that runs simple background tasks in Claude Code on a cheaper Claude model. It never switches upward, never touches demanding work, stays hands-off when it cannot tell what a task needs, and one command turns it off. Every decision about your accounts and plans stays yours.
Help us build the right thing.
2 minutes. 8 questions. Directly shapes the next version.
Help us build the right thing
Take back control of your AI stack.
See your usage, limits and spend across every AI you pay for.
Mac app and web dashboard. Windows coming this summer.