I built this because I needed it.

The story

I spend around $300 to $400 a month on AI tools.

Claude, ChatGPT, Cursor, Perplexity, sometimes more. Some are for writing. Some are for code. Some are for research. Some are for thinking through product decisions. Each one has its own limits, reset windows, hidden quotas, slow requests, model-specific caps, and billing logic.

At some point, the problem stopped being "which AI tool should I use?"

The real problem became:

What do I have left, and what should I use next?

One afternoon, in the middle of real work, my main AI tool hit a limit. No warning. No useful signal. No clear path forward. I had context in my head, work in progress, and no idea which tool to switch to without wasting time or burning another quota.

So I built claudeKarma first: a free Chrome extension that tracks Claude usage directly in the toolbar.

It reached thousands of users without much promotion. Developers, writers, lawyers, recruiters, translators, researchers, consultants. The feedback was clear:

• "Can you do this for ChatGPT?"
• "Can you do this for Cursor?"
• "Can you track everything in one place?"

tokenkarma is that place.

Who built this

I am Jean-Rémi Larcelet-Prost, an indie maker based in France.

I have spent around ten years building products across web, mobile, automation, and infrastructure. I built claudeKarma as a free extension because I needed it myself. tokenkarma is the paid Mac app that came from the same pain, but for the whole AI stack.

You can reach me at [email protected] or on LinkedIn.

What tokenkarma does

tokenkarma is a Mac menu bar app for heavy AI users.

It helps you:

• see how much usage you have left across your AI tools;
• avoid discovering limits in the middle of work;
• understand which provider is close to a cap;
• choose the right model or provider before starting a task;
• stop wasting your strongest quota on work that a cheaper or faster model could handle.

The first version focuses on usage tracking, alerts, and practical recommendations.

The long-term goal is bigger: help people route work across AI tools without losing momentum, context, or money.

Why this exists

AI providers are not built to help you manage your AI stack.

Their dashboards show parts of your usage, but they are not designed to help you decide whether you are wasting a subscription, whether you should switch models, or whether you are about to burn your best quota on the wrong task.

tokenkarma is built from the other side.

It is built for the person paying the bills.

The goal is not to make you use more AI. The goal is to help you use the right AI, at the right time, with less waste.

What tokenkarma will never do

tokenkarma will never read the content of your prompts or AI responses.

It will never ask for your AI provider passwords.

It will never store your provider cookies on tokenkarma servers.

It will never pretend to be affiliated with the AI providers it tracks.

The Mac app uses your existing browser sessions and local usage signals to help you understand your own usage. Provider cookies stay on your Mac, encrypted in macOS Keychain.

When tokenkarma needs server-side systems, such as account access, payments, provider configuration, or future cloud features, those systems are separated from your provider sessions.

Founders access

Founders access exists to help fund the first public version while keeping tokenkarma small, independent, and focused.

Founders get lifetime access to tokenkarma Core.

That means the Mac app, local usage tracking, provider alerts, local history, and core app updates.

Founders also get the first months of AI Monitor included.

AI Monitor is the AI-powered layer that helps analyze your task and recommend which model, provider, or workflow to use. Because it uses cloud AI credits and ongoing infrastructure, it is not part of lifetime Core access forever.

After the included period, continued access to AI Monitor will require a subscription.

The distinction is simple:

Founders get tokenkarma Core for life. AI-powered cloud features are subscription-based.

What we may ask in return

tokenkarma will work best if users help shape it.

With explicit consent, we may collect limited product signals to understand what works, what breaks, and what should be improved: provider status, timestamps, model identifiers, token counts, error types, and feature usage.

We do not need your prompt content to build this product, and we do not want it.

You will be able to opt out, delete local data, and keep your Founder access regardless.

We may also send short product emails asking specific questions. Reply if you want. Ignore them if you do not.

Open source heritage

claudeKarma started as a free, open-source extension because I wanted to solve a specific problem in the simplest possible way.

That extension proved the pain was real.

The Mac app, backend, payment system, and future cloud features are closed source so tokenkarma can become a sustainable product.

The principle stays the same:

Build useful tools. Explain what they do. Keep the trust.

Roadmap

The first public version focuses on the essentials:

• Mac menu bar usage tracking;
• the first supported AI providers;
• alerts before limits become a problem;
• AI Monitor recommendations;
• Founders access;
• fast provider fixes when integrations change.

After that, the roadmap will be shaped by real usage and Founder feedback.

Likely next areas include more provider integrations, better model recommendations, cloud sync, historical analysis, and smoother switching between AI tools.

No fake certainty. No giant roadmap theater. The product will grow from the problems users actually hit.

Contact

For everything tokenkarma: [email protected]

For privacy questions: [email protected]

For partnerships, press, or creator outreach: [email protected]

Plain subject lines win.