Evidence describes a failure. Someone still has to rebuild the actions, state, dependencies, timing, and environment that caused it, and hope the rebuild is faithful.
A repro is an artifact, not a narrative. It runs before the change, runs after the change, and answers one question without discussion: did the same failure happen again?
The exact sequence of actions, inputs, and timing that starts the failure, reduced to only what matters.
Which databases, services, files, devices, and external systems are inside the failure, and which are noise.
The clocks, seeds, schedules, versions, and conditions that must be held still for the failure to return.
The machine-checkable condition that proves the same failure occurred. Not a screenshot. Not a judgement call.
From production through an SDK, from a failing CI job, or from a run you point at.
It is replayed against the same failure contract. If it does not happen again, it is not a repro.
Irrelevant actions, inputs, and dependencies are removed until only the cause remains.
What is left becomes a stable bug ID you can paste into a PR.
A developer, an agent, or CI runs the same failure on demand.
The fix is judged by the same artifact that proved the bug.
The same artifact defines the bug and the definition of done. Nothing about the fix is a matter of opinion.
The failure is captured with its trigger, boundary, envelope, and oracle. Reconstructing it again is never necessary.
A held guarantee is quiet by design. The proof that it still holds is that the repro runs and still fails to reproduce.
The class of defect becomes structurally impossible to ship again, not merely absent from today's build.
Derived crash signatures, replay steps, and triage state. Never source code, simulators, or builds.
Your runners check out your code and replay the failure. Only the verdict comes back.
Structural fingerprints of inputs: length, charset, script. Never the values your users typed.
Run the fuzzing on your own machine and keep your repros, buckets, and evidence in one hosted place.
Private managed service for small teams putting production bugs into Repro It.
For shipping teams debugging real production failures.
For larger orgs with more apps, seats, and retention.
For high-volume and regulated teams.