Feature flags per environment with percentage rollouts, evaluated inside your app so a flag check is never a network call.
On in development, off in production, or a percentage in between, without a deploy and without a config file somebody forgets. Each environment has its own key, so switching something on in staging cannot switch it on anywhere else.
The client fetches the rules once and decides locally, so a check costs nothing in a hot path and our outage is not your outage. A flag nobody has defined is off, so a typo fails closed.
Every change records what the flag was as well as what it became. After an incident nobody asks what a flag is set to now; they ask what it was at the time.
Named rather than left out.
Percentage rollouts and on or off. No targeting by attribute, country or cohort.
Monitor knows the crash free rate per release and Flags does not ask it yet, which is the reason to have built this here.
There is no way to pay for anything on this platform.
Flags shares an account, a project and a bill with Monitor, Analytics and the rest. Two tools that already agree on what a release is beat two tools you reconcile by hand.