Skip to main content
Rollback moves the LIVE pointer to an earlier deploy. New runs use the older version; in-flight runs finish on whatever they started with. Rollback is:
  • Instant a pointer update, not a redeploy. Seconds.
  • Atomic the whole Flow rolls at once. No partial state.
  • Reversible you can roll forward again, or to any other version.
  • Recorded no version is created or destroyed. The LIVE badge moves; the version list stays intact.

​
When to roll back

  • Signals spike right after a deploy.
  • A regression slipped past simulation and was caught late.
  • The proper fix takes time and you need to buy some.
The cost of rolling back is very low. Roll back sooner rather than later.

​
How to roll back

Versions → click the version to return to → Roll back to this deploy → Confirm rollback. The button turns into a [Confirm] button / [Cancel] button pair so a stray click can’t roll anything back. The version currently marked LIVE has no rollback button: it’s already what’s serving. To roll back from the CLI, see flows in the CLI.

​
What rollback leaves untouched

  • The Draft your working canvas is not modified. To build on the rolled-back version, use Restore to Draft.
  • Version history the version you rolled away from stays in the list, and you can roll back to it anytime.
  • Traces and signals run records stay attached to the version that ran.

​
Rollback vs. restore

As the version detail panel puts it: rollback moves the LIVE pointer your external API triggers use, while restore overwrites the canvas’s working draft with that version.
  • Rollback production changes. Draft untouched.
  • Restore Draft changes. Production untouched.
Use rollback to stop the bleeding; use restore when you want to hand-edit an older version into a new fix.

​
Rolling back an Improvement

If you know which Improvement caused the issue, roll back just that improvement instead of the whole version: find it under Versions → Improvements and use its one-click rollback.