SDK request usage
Commercial counts the requests your apps make for flags, so you can see how close you are to your plan's allowance. It reports usage. It never cuts your traffic.
What counts#
A full answer (HTTP 200) from one of the three delivery endpoints: evaluation.server, evaluation.client and evaluation.stream.
| Reply | Description |
|---|---|
| 200 from a delivery endpoint | Counts. |
| 304 Not Modified | Free. An SDK that is idle and polling every 30 seconds costs nothing. |
| Failures | Free. |
| SDK error reports and analytics calls | Free. |
Why 304s matter
Client SDKs send the tag from their last answer with each poll (If-None-Match, on both GET and POST), so an unchanged configuration costs a free 304 instead of a billable 200. The JavaScript family, Flutter and Swift do this only for flags they fetched themselves for the same person.
When the count resets#
The period starts on the day of the month the organization was created (UTC, at midnight) and runs to the same day the next month: created on the 14th, counted 14th to 14th. A 29th, 30th or 31st becomes the last day of a shorter month. The Plan and usage page shows when it resets. It is not tied to when a payment was made.
What happens at the limit#
- Nothing is cut off. Flags keep being served.
- At 80% and at 100% of the allowance, the people who can edit organization settings get one email each per period.
- On Pro and Business the overage is estimated on the plan page. It is not charged yet.
Counts are kept in memory per server and written every 30 seconds, so a crash loses at most 30 seconds of counting.
Not available yet#
Slowing a Free key that stays far above its allowance is not built. If it is, it will slow refreshes only: SDKs keep the last good flags and honor Retry-After, and traffic is never refused outright.