Every plan includes the full rehearsal engine, independent verification and the deterministic release gate.
One run of the eleven-step workflow for one change: hypotheses, experiments, remediation, verification, packaging, staging and a policy decision. Cancelled runs don't count.
No. Agents propose hypotheses and fixes; a deterministic, digest-pinned policy engine makes the decision from recorded evidence. Missing evidence is never treated as a pass.
Yes. FutureFault is MIT licensed and ships as a single container. Run it on your laptop, in your cluster, or on Google Cloud Run with the included deploy script.
Limits are enforced by the API. The operator assigns a plan per workspace with a single CLI command; billing isn't wired into the open-source build.
New rehearsals return a clear 402 message and the sidebar meter turns amber at 80%. Existing evidence stays readable and exportable.