Usage and alerts
feat meters one thing: how often your SDKs call the edge. Seats, flags, environments, projects, and the number of people using your product are not measured and do not affect your bill.
What counts as a request
Section titled “What counts as a request”| Call | Counts |
|---|---|
| A server SDK fetching the datafile | Once, including a 304 Not Modified |
| A client SDK evaluating remotely | Once per evaluation request |
| An SDK opening a stream | Once, at connect |
All of these weigh the same. Two things are worth calling out:
- Updates feat pushes down an open stream are never counted. Changing a flag does not cost you anything, however many SDKs are listening.
- A
304costs the same as a full response. A server SDK polling an unchanged datafile is cheap for you in bandwidth, but it is still a request.
Streaming is the cheaper path by a wide margin. A streaming server SDK spends roughly 4,400 requests a month per instance: one connect plus a safety-net poll every ten minutes. The same SDK on a 30-second poll loop spends about 86,400.
A browser or mobile session costs more than one request. Starting the SDK evaluates once, subscribing opens a stream, and the background poll keeps ticking slowly while the tab is visible, so budget two or three per session and more for a long one. setContext re-evaluates, so a login costs one more.
Seeing your usage
Section titled “Seeing your usage”Open Organization settings then Usage. It shows requests used in the current billing period against your plan’s allowance.
The billing period is anchored to the day you subscribed, not the calendar month. Subscribe on the 14th and the period runs 14th to 14th. Months too short for the anchor clamp for that period only, so the 31st becomes the 28th in February and returns to the 31st afterwards.
At the allowance
Section titled “At the allowance”Your flags keep evaluating either way. Production is not the place to enforce a billing limit, so nothing is cut off when you cross the line.
What differs is the bill:
- Paid plans keep going and the extra is billed at the published usage rate. It falls again when your traffic falls.
- Free has no card on file and no overage rate, so there is nothing to pay. If your usage settles above the allowance, a paid plan is how to get more headroom.
Alerts
Section titled “Alerts”Every organization gets email alerts as it approaches its allowance. The defaults are 50%, 80%, 95%, and 100%.
Alerts go to every Admin in the organization. Each level fires at most once per billing period, so crossing 80% does not re-notify you every day you stay above it. If usage jumps past several levels at once, feat sends one email for the highest level reached rather than one per level.
100% is always on. It is the notification that your allowance is gone, and it cannot be added, edited, or removed.
Configuring them
Section titled “Configuring them”Open Organization settings then Usage alerts. You need the Admin role to change them.
| Configurable levels | Up to 4, in addition to the permanent 100% |
| Valid range | 10 to 99. 100 is the permanent one and cannot be added |
| Duplicates | Rejected. One alert per level |
Removing a level stops future emails at it. It does not retroactively un-send one already delivered in the current period.
Related
Section titled “Related”- Projects and organizations for where organization settings live.
- Members and roles for who holds the Admin role.
- Evaluation model for which SDKs call the edge and when.
- Pricing for current allowances and rates.