Forks and branches

Use a fork for several independent, concurrent, multi-step routes:

<fork id="distribution" join="instagram facebook">
  <branch id="tiktok">
    <step id="prepareTikTok"><script>return prepareTikTok();</script></step>
    <step id="publishTikTok"><script>return publish(context.steps.prepareTikTok);</script></step>
  </branch>

  <branch id="instagram">
    <step id="publishInstagram"><script>return publishInstagram();</script></step>
  </branch>

  <branch id="facebook">
    <step id="publishFacebook"><script>return publishFacebook();</script></step>
  </branch>
</fork>

<fork> attributes:

<branch> requires id and optionally accepts name and description. It must be non-empty and may contain multiple steps, choices, switches, parallel groups, and approvals.

Execution rules:

• Branches begin concurrently; each branch remains sequential internally.
• A branch sees context available before the fork and earlier outputs in that branch, never sibling outputs.
• The main continuation waits only for selected joined branches.
• Only joined branch outputs become visible to the continuation. Completion timing never makes an unjoined output visible.
• join="none" releases the continuation without waiting.
• Even after the continuation is released, the workflow itself does not report success until every fork-owned branch settles.
• A failure in a joined branch blocks the continuation. An unjoined failure does not block it, but the overall workflow still settles as failed after all branches finish.
• Nested forks anywhere inside a fork-owned branch subtree are rejected in v1.
• A terminal fork preserves the last earlier main-route value. A workflow made only of a result-less fork is invalid.