RecoverDocumentation
Reflects the code as of 28 July 2026. If the product changes and this page does not, that is a bug worth emailing about.
These pages describe what the software does, not what it aspires to do. Where a behaviour is a default you can expect to change, it says so. Where something is missing, it says that too.
How it works
The whole path from a failed payment to a recovered one, in order.
Connecting Stripe
What the OAuth connection grants, what it cannot do, and how to revoke it.
Decline classification
Exactly which codes are retried, which are not, and why.
Retry scheduling
The actual schedule, including the cases where it schedules nothing.
Dunning emails
The sequence, the timing, and what the emails currently do not do.
Attribution and billing
How a recovery is credited, and when you are not charged.
The 90-day backtest
The scan that runs at connect time, and how to read its estimate.
A note on these docs
RecoverFlow is early and run by one person. These pages are written from the source, and several of them end with a list of things the product does not do yet. That is deliberate. Finding out after you connect is worse for both of us.
If you want the conceptual background rather than the implementation, the guides cover Stripe's own behaviour in more depth, with citations.