Error handling and branching

What you will do

Add error handling and conditional branching to make your workflow resilient.

Why error handling matters

Workflows run unattended. If a node fails (the API is down, the model returns an unexpected response, the input is malformed), the workflow stops silently unless you handle the error. In production, a silent failure means missed reports, unsent messages, and data that never gets processed.

Add an error path

Every node in n8n has an error output. By default, errors stop the workflow. To handle them:

1. Click the Settings tab on a node.
2. Enable Continue on Fail. The node will pass the error information to the next node instead of stopping.
3. Add an IF node after the failing node. Check whether the output contains an error. If yes, route to an error-handling path (send a Slack notification, log the error, retry).

Add conditional branching

Use the IF node or Switch node to route the workflow based on conditions.

Example: After the AI Agent classifies an email, use a Switch node to route:

• "bug" → Create a Jira ticket
• "billing" → Forward to the finance team's Slack channel
• "spam" → Archive and do nothing

The Switch node checks the value of a variable and sends the data down the matching path.

Retry logic

For transient failures (API timeout, rate limit), add a retry loop:

1. After a node that might fail, add an IF node checking for an error.
2. If there is an error, route back to the original node.
3. Add a Wait node before the retry to avoid hammering the API.
4. Add a counter variable that stops retrying after 3 attempts and routes to the error notification path instead.

Test it

Deliberately cause a failure (use an invalid API key, send malformed input) and check that the error-handling path runs correctly. Verify that the error notification arrives where you expect.

What you should see

A workflow that handles failures gracefully: errors are caught, logged, and reported instead of silently stopping the workflow.