Handling Error Scenarios

Error handling is the safety net of a workflow. It does not run in the normal flow and activates only when something unexpected happens. There are three kinds of unexpected events a workflow may encounter:

Invalid Input: The customer typed something that an input node could not parse, and all retry attempts are exhausted.

Error: An API call or action node failed (for example, the transfer system was unreachable, or an email connector failed).

Timeout: An API operation took too long to complete.

Error handling is a pattern, not a separate node type. Each relevant node has dedicated output handles for these events. You connect those handles to whatever node you want to handle the failure; typically a Statement node that apologizes and either retries, branches, or leads into a Chat Transfer.

Handle	Appears on	When it fires
On Invalid Input	All input nodes	Customer's response is invalid after all retry attempts.
On Error	API nodes, all Action nodes	A call or action fails at runtime.
On Timeout	API nodes only	The API operation exceeds the timeout limit.

Best practices:

Always connect error handles. An unwired error handle means the workflow has no graceful path for that failure, and the customer may be left without feedback.
For API nodes, connect both On Error and On Timeout to meaningful recovery paths.
For action nodes (Chat Transfer, Send to AI Agent, Send Email, Trigger Call), connect On Error to a Statement that informs the customer and offers an alternative.