I cancelled an AI subscription after seeing my actual usage data for the first time
I was paying for 5 AI subscriptions: Claude Pro, Cursor Pro, ChatGPT Plus, Copilot, and Gemini Advanced. $90/month total.
I assumed I was using all of them heavily. Turns out I was wrong.
When I built TokenBar (a menu bar app that tracks AI usage limits), the first thing I noticed was that my Gemini Advanced usage was near zero. I'd signed up for it, used it twice, and forgot about it. $20/month for 3 months = $60 wasted.
But the bigger insight was about my other tools:
- Claude: I was hitting 80%+ of my limit almost every day. Worth every penny.
- Cursor: Using about 60% of fast requests. Solid value.
- ChatGPT: Maybe 20% usage. I was using it mostly as a Claude backup when rate-limited.
- Copilot: Under 10%. VS Code autocomplete was nice but I wasn't really dependent on it.
Without usage data, all subscriptions feel equally necessary. With data, the picture is obvious.
I cancelled Gemini ($20/mo saved) and I'm evaluating whether Copilot is worth keeping. That's potentially $30/mo saved — $360/year.
TokenBar paid for itself in the first week.
What it shows:
- Usage vs. limit for each AI tool
- Reset windows (so you can plan heavy work)
- Burn rate (are you pacing well?)
- Which tools still have capacity
20+ providers supported: Claude, Cursor, ChatGPT, Copilot, Gemini, OpenRouter, Vertex AI, Augment, Amp, JetBrains AI, and more.
$4.99 one-time. Not a subscription.
Replies