Hi @projetosaberebemesta Welcome!
The 5-minute figure is outdated. Cloud does not enforce a short execution cap on Trial or Starter, so a run with several sequential API and LLM calls that goes past five minutes will still finish.
See this:
Timeouts are opt-in per workflow: Settings > Timeout Workflow > Timeout After. Cloud caps the maximum value you can pick per plan, and you cannot raise it beyond that on Cloud because env vars are not exposed there. EXECUTIONS_TIMEOUT and EXECUTIONS_TIMEOUT_MAX only exist on self-hosted. For the exact ceiling on your plan in writing, ask help@n8n.io.
Each workflow carries its own timeout setting, and a workflow called by the Execute Sub-workflow node runs as its own execution, so the sub-workflow’s timeout applies to the sub-workflow while the parent’s clock keeps running the whole time it waits.
When a timeout is reached, n8n cancels the execution once the current node finishes. It is marked canceled and nothing resumes automatically.
The real constraint on Trial and Starter is memory, not time: both are 320MiB RAM. Long editorial chains die there first, so push the heavy steps into sub-workflows that return only small results.
Plan-by-plan limits if you’re weighing Starter against Pro: