Every feat plan has unlimited seats, including the free one. This is not a promotion. It’s the pricing model, and it exists because seat pricing is wrong for this category in a way that compounds as your team grows.
Who actually needs to see your flags
A feature flag is a shared fact about production. The engineer who wrote the flag needs it, obviously. But the person who most often needs to look at it is someone else: the support engineer triaging “why does this customer see the old checkout,” the PM deciding whether the rollout is far enough along to announce, the on-call engineer at 2 a.m. who didn’t write the flag and wants to know what flipping it off will do.
Under seat pricing, each of those people is a line item. So teams do what teams always do: they don’t buy the seat. The PM asks an engineer to check. The support engineer pastes screenshots into Slack. The one shared login gets passed around, which your auditor will eventually have opinions about. The tool that was supposed to make production legible to the whole company becomes legible to the eight people whose seats were approved.
We’d rather your whole company log in. The audit log is more useful when actions are attributable to real people, approvals mean more when the approver is the actual reviewer, and a kill switch is safer when the person who notices the fire doesn’t have to find someone with access.
What our costs actually scale with
The honest version of any pricing model is the vendor’s cost curve with a margin on it. Ours barely moves when you add a teammate: another login costs us a session row. What moves it is traffic, specifically how often your SDKs ask our edge for something.
So that is the meter, and it is the only one. One request is one call from an SDK to us: a server fetching the datafile, a browser or mobile SDK evaluating remotely, a stream connecting. Free covers a million a month, Pro ten million at $29, Growth twenty-five million at $99, and past the allowance it is a flat $6 per extra million rather than a cliff. Connecting a service costs nothing by itself, because charging per connected process taxes teams for having a normal microservice architecture.
Note what is not in there: how many people use your product. We used to meter that, and we were wrong to. A backend fleet makes about the same number of requests whether ten people or ten million are on the other side of it, so a good month for you should not be an invoice event. The one thing we never count is the updates we push down an open stream when you change a flag. A meter that charges you for changing flags is a strange thing to sell on a flag platform.
The failure mode we’re avoiding
Per-seat flag tools produce a predictable culture, and if you have worked near one you have seen it: flags become an engineering-only concern, rollout state lives in one team’s heads, and every incident review includes ten minutes of reconstructing who changed what because half the actions trace to a shared account.
Unlimited seats costs us almost nothing and buys your team the version of feat where everyone can see what’s live. That trade isn’t close.
If the arithmetic makes another vendor cheaper for your shape of product, run it and pick them; the comparison we published shows ours, and the calculator will do your side of it. But nobody on your team should cost $15 a month to look.