Escriba ecosystem · Open source · MIT

All your AI limits, in one place

Claude, ChatGPT, Gemini, Grok and Perplexity cut you off when you least expect it. TokenYou shows how much of each window you've used, when it resets, and warns you before you hit the limit — without ever reading your conversations.

Install from GitHub View the code Coming soon to the Chrome Web Store · Manual install available today
0
sites accessed on install — each platform is enabled one by one
0
code run inside your pages, on any platform, ever
0
telemetry, analytics, own servers or dependencies
2
total manifest permissions: storage and alarms
What it does

Your instrument panel for AI

Every platform has its own windows — 5-hour sessions, weekly caps, credits, remaining searches. TokenYou normalizes them into a single panel you read at a glance.

Real meters, not estimates

Reads the same quota data each platform shows its own interface: exact percentage used and reset time, per window and per model.

A countdown to every reset

"Resets in 2 h 05 m" next to each bar. The toolbar badge shows your worst meter in its color — green, amber or red — without opening anything.

Show only what matters

Hide the meters you don't use with one click. Hidden ones drop out of the panel and the badge, so the number you see is always the one that affects you.

Claude Code included, free

Anthropic's limit pool is shared: the same claude.ai meter also reflects what you spend from Claude Code and the desktop app.

Add your own services

Use Cursor, Copilot or another tool with a quota? Define it as JSON in 30 seconds, without touching the extension's code, and it shows up as one more platform.

Genuinely lightweight

~30 KB of code, zero frameworks, zero build. The service worker sleeps between refreshes: near-zero CPU and memory use.

Security

Built so you don't have to trust us

Usage trackers sit next to your most sensitive browser data: your AI conversations and your logged-in sessions. In 2025, "AI" extensions with hundreds of thousands of users were caught exfiltrating whole conversations. TokenYou was built against that scenario.

Typical trackerTokenYou
Access on installAll target sitesNone
Cookies permissionCommonNever
Code on your pagesInterceptors / scrapingZero, on all
Your conversationsVisible to the extensionUnreachable by design
TelemetryAnalytics, heartbeatsZero third-party traffic
SourceClosedMIT + reproducible build

Architectural incapability

TokenYou doesn't "promise" not to read your chats: it can't. No script runs inside any page. It only makes background fetches to each platform's quota endpoint and reads numbers.

Your session is never touched

Requests are same-origin: the browser attaches the session on its own. The extension has no cookies permission and can't see the credential.

Per-platform, revocable

Freshly installed, TokenYou accesses no site. You enable Claude, ChatGPT or whichever you use, one at a time, and can remove any of them whenever you want.

Verifiable, not just auditable

The package is built with git archive from a tagged commit: same commit, same zip, published hash. Anyone can confirm what's installed is exactly this code.

Platforms

Five out of the box, infinite by definition

Each platform is an isolated adapter calibrated against real responses. If one changes its internal API, only that platform degrades to "unavailable" — the rest keeps working.

Claude

5-hour session, weekly cap (global and per-model), and extra-usage spend in dollars. Also covers Claude Code and the desktop app.

/api/organizations/{org}/usage

ChatGPT

5-hour and weekly windows with exact percentage and reset time, plus the detected plan (Free / Plus / Pro).

/backend-api/wham/usage

Gemini

Current and weekly compute usage — the same real figure Google's usage page shows, read from its internal RPC.

batchexecute · jSf9Qc

Grok

Remaining queries out of the total for each rolling window, by request type.

/rest/rate-limits

Perplexity

Remaining Pro, Research, Labs and agentic searches on your plan.

/rest/rate-limit/all

Yours

Cursor, GitHub Copilot, internal dashboards: any service with a JSON usage endpoint is added from the options page.

custom services →
How to read the meters

Green, amber, red — and two marks on every bar

Every meter uses the same semantic scale, and each bar carries two fixed threshold marks (at 60% and 85%) so you see how far you are from the next zone. The toolbar badge takes the color of your worst visible meter.

0 – 59% · green

Comfort zone. Use freely: you have plenty of headroom in this window.

60 – 84% · amber

Caution zone. Worth pacing yourself, or checking the reset before a long task.

85 – 100% · red

Critical zone. You're about to hit the limit: save what matters or switch platforms.

Custom services

Your tool isn't listed. Add it in 30 seconds.

For advanced users: define any service with a JSON usage endpoint as a declarative entry — the origin, the URL (mandatorily on the same host) and dot-paths to the numbers. The Test button requests the permission and shows exactly what was read before you trust the meter.

{
  "id": "copilot",
  "name": "GitHub Copilot",
  "origin": "https://github.com/*",
  "url": "https://github.com/settings/billing/copilot_usage_card?…&unit_type=1",
  "meters": [{
    "id": "premium",
    "label": "Premium requests",
    "usedPath": "netQuantity",
    "totalPath": "userPremiumRequestEntitlement"
  }]
}
FAQ

Frequently asked questions

Can TokenYou read my conversations?

No — and not by a promise but by architecture: no TokenYou code runs inside any page. It only makes background requests to each platform's quota endpoint — the same one that platform uses to show you your own usage — and reads numbers: percentages, remaining counts and reset times.

Why does it ask for each site separately?

Principle of least privilege: freshly installed, TokenYou has access to no site. Each platform you enable is an individual permission Chrome shows you, and that you can revoke at any time with the "Remove" button. If you never enable a platform, the extension can never touch it.

Does it send any data to a server?

There is no exfiltration path: zero telemetry, zero analytics, zero own servers, zero external dependencies. The only network traffic is to the platforms you enabled, and only to their quota endpoints. Everything is stored in your browser's storage.local.

How do I verify the published build is this code?

The package is built with git archive from a tagged commit, which makes it byte-reproducible: clone the repo, run the same command and compare the SHA-256 hash with the one published on each release. GitHub CI generates it automatically.

What happens when a platform changes its internal API?

Quota endpoints aren't public contracts and can change. Each platform lives in an isolated adapter: if one breaks, that card shows "unavailable" and the rest keeps working. Being open source, the fix usually lands within days — and you can report or contribute it yourself on GitHub.

Does it track Claude Code, Cursor or Copilot?

Claude Code yes, automatically: it shares the limit pool with claude.ai, so the Claude meter already includes it. Cursor and GitHub Copilot are added as custom services from the options page (templates are provided), reading their web dashboards with your session.

Is it free? What's the business model?

Free and MIT. There's no hidden business model: no data is sold (there is no data), no premium, no accounts. It's community software, part of the Escriba ecosystem.

Does it help bypass platform limits?

No. TokenYou is a read-only monitor of your own quota — it shows you what the platforms already show you, better organized. It does not bypass, extend or manipulate any limit, in line with Chrome Web Store policy.

About

Part of the Escriba ecosystem

TokenYou is part of the Escriba ecosystem, the family of open tools built on one conviction: software that lives next to your most sensitive data should be verifiable, not merely trustworthy. Every piece of the ecosystem is open source, processes locally everything it can, and treats permissions as something earned, not assumed.

TokenYou was born from a simple frustration — hitting the Claude limit halfway through an idea — and an uncomfortable observation: the extensions that solved it asked for more access than they needed, closed-source, right next to the most valuable conversations we have with an AI. The answer was to build the meter we wanted to use: one that couldn't spy even if it tried.

Discover the rest of the suite at getescriba.com.