BYOK: Why We Don't Pay for Your AI Tokens
Most SaaS companies hide AI costs in your subscription. We hand you the bill directly. Here's why that's better for everyone.
Every AI-powered SaaS tool faces the same dilemma: who pays for the tokens?
Most companies absorb the cost and bake it into your subscription. This creates a perverse incentive: the more value you get from the AI, the more money they lose. So they throttle. They rate-limit. They give you a watered-down model.
Our Approach: Bring Your Own Key
Grumpy uses a BYOK (Bring Your Own Key) model. You plug in your own OpenAI API key, and we use it directly. No markup. No throttling. No surprises.
Why This Is Better
- Full transparency — You see exactly what each analysis costs on your OpenAI dashboard
- No artificial limits — Analyze as many errors as you want
- Model choice — When GPT-5 drops, you get it immediately
- Cost control — Set your own OpenAI spending limits
The Math
A typical error analysis uses ~2,000 tokens. At GPT-4o rates, that's roughly $0.01 per error. If you're throwing 100 errors a day (and if you are, you need Grumpy), that's $1/day in AI costs.
Compare that to enterprise error tracking tools charging $29-99/user/month with inferior analysis.
Setup
Navigate to Dashboard → Settings and paste your OpenAI API key. We store it encrypted at rest with AES-256. We never see your key in plaintext after initial encryption.
That's it. Full AI-powered error analysis for the cost of a few pennies per crash.