What to do when an n8n workflow fails: errors and alerts

It is 9 in the morning and a customer writes to you: "I filled out the form on Monday and nobody has answered me." You check and find that the workflow that passes leads to the CRM has been broken for three days, because a credential changed. Nobody received an alert, so nobody knew.

That is what happens when you automate without monitoring. The good news: in n8n, getting alerted when something fails takes minutes to set up.

Why does an n8n workflow fail without anyone noticing?

Because, by default, an error in an automation is only recorded inside n8n, and nobody checks that log every day. An expired credential, an API that changes or a field that arrives empty is enough to stop the workflow.

The most common failures in real businesses are these:

• A credential or token that expires.
• An external service that does not respond.
• A required field that arrives empty (a phone without a number, an email without a valid format).

If you want to review how a workflow is built from scratch, start with what n8n is and how it works.

What is an error workflow in n8n?

It is an independent workflow that starts with the Error Trigger node and runs only when another linked workflow fails. According to the official n8n documentation, that node receives the details of the workflow that failed and of the errors, and from there it runs whatever you define.

Three details from the same documentation worth knowing:

Point	What the documentation says
Publishing	Error workflows do not need to be published
Default value	A workflow with this node uses itself as its error handler
Testing	It cannot be tested manually: it only fires when a workflow fails automatically

Common mistake: waiting to "test" the alert by pressing the run button. Since the Error Trigger only fires on a real failure, the right test is to cause an error on purpose in a test workflow, for example with the node we explain below.

What is the Stop And Error node for?

It is used to make a workflow fail on purpose, with a clear message, when a condition is not met. The n8n documentation describes it as the node to display custom error messages, make an execution fail under certain conditions and send error information to error workflows.

It has two operations: Error Message, where you write the error text, and Error Object, where you throw a JSON object with the properties you want.

A practical example: if a lead arrives without a phone number, instead of letting the WhatsApp step fail in a confusing way, you stop the workflow with the message "Lead without phone, check form source". The alert arrives with a cause anyone can understand.

How do I build the alert step by step?

Create a new workflow with the Error Trigger, connect it to a channel your team does check and link it to the workflows you want to monitor. The recommended order:

1. Create the error workflow with the Error Trigger node as the first step.
2. Add an alert step: a WhatsApp message or an internal email to the person in charge.
3. Include in the message the workflow name, the step that failed and the time.
4. Link that error workflow to every important automation (sales, payments, support).
5. Cause a failure on purpose in a test workflow to confirm the alert arrives.

If your workflows move leads into a CRM, the alert is part of the processes you can safely automate with AI, because it turns an invisible failure into a task with an owner.

What changed in n8n this week?

n8n published several maintenance releases in early October. According to its releases on GitHub, version 2.42.3 (October 5, 2026) includes a fix to keep the task runner running after an unhandled promise rejection, and version 2.43.0 (October 6, 2026), marked as a pre-release, blocks publishing workflows that have unconnected required inputs on a node.

None of them replaces what came before: they are improvements that reduce failures, but they do not remove the need for an alert. If a third party runs your n8n, ask which version it runs and when it gets updated.

What else should you watch besides errors?

The failures that do not generate an error. A workflow can finish "fine" and still not do what you expect, for example if traffic stops arriving. Complement the alert with a weekly check of how many leads entered the CRM and compare them with your ads. If you use AI agents, also define clear instructions for your agent and the cases where it must hand the conversation to a person, as explained in the guide to the AI agent for WhatsApp.

In short

An automation without error alerts is a gamble. The Monday lead nobody answered is avoided with a short error workflow, a clear message and a channel your team checks. If you would rather not build or maintain it yourself, learn about our n8n automation service.