Credits & overage
How AI usage is metered and what happens when you exceed your monthly allotment.
The work that costs us money is the AI analysis — diff summaries, breaking change detection, AI-fallback parsing for sources without a stable adapter. We meter that work in credits.
What burns a credit #
- Diff analysis on a new release. Each release we summarize for a tracked source consumes credits proportional to the diff size. Most releases are 1 credit; very large diffs may consume more.
- AI-fallback parsing. Polling a source that doesn't have a stable parser burns a small amount per poll — except when the page hasn't changed (we cache by content hash and skip the LLM call).
- MCP / API calls that trigger fresh analysis. Reads of cached data don't cost credits.
What doesn't cost credits: dashboard navigation, viewing past summaries, acknowledging alerts, lockfile uploads (parsing is local), polling sources with stable parsers (no LLM involved).
Plan allotments #
Each tier includes a monthly allotment of credits sized to typical usage at that tier. Most users never hit the cap — the metering exists so that one heavy outlier doesn't make the service unviable for everyone else.
Overage #
If you exhaust your monthly allotment before the period ends:
- Soft cap. Polling continues but new diff summaries pause until you top up or the next billing period rolls over.
- Top-ups. You can purchase a one-time credit pack from Settings → Subscription. Top-up credits don't expire and stack with the monthly allotment.
- No surprise bills. We don't auto-charge for overage — usage above your plan stops the AI side cleanly rather than ringing up a bill.
Tracking your usage #
The settings page shows credits used vs. allotted for the current period plus a 7-day usage trend. If the trend looks like you'll exhaust the allotment, you'll see a banner suggesting an upgrade or top-up.