n8n error handling made simple: alerts, retries and an error workflow

Building an automation in n8n is the fun part. What really decides whether it saves you time is something else: what happens the day it fails. An expired key, an API that doesn't answer, a message in an odd format… and your agent stops replying without anyone noticing.

In this guide you'll see, without coding, the three layers every workflow that works for your business should have: retry, decide what to do when a step fails, and alert you when something breaks.

Why silent failures are the biggest risk

A workflow that errors and tells you is a small problem: you fix it in ten minutes. A workflow that fails without telling you is a big one: customers left without answers, bookings that never get saved, or invoices that don't reach your spreadsheet for days.

The most common failures in small-business workflows are:

An external service doesn't answer for a moment (WhatsApp, Google, the AI model). Retrying usually fixes it.
Expired credentials or revoked permissions. These don't fix themselves: you need to know.
Unexpected data: an empty field, an email with no subject, a phone number in another format.
Usage limits from your AI provider or an API.

Layer 1: automatic retries on the node

Many errors are temporary. For those, n8n has an option on every node: open the node, go to Settings and turn on Retry On Fail. You'll see two values:

Max Tries: how many times it tries before giving up.
Wait Between Tries: how long it waits between attempts (in milliseconds).

Use it mainly on nodes that call external services: the AI model, HTTP requests, Google Sheets or message sending. A sensible starting point is 3 tries with a few seconds between them. Think twice before using it on nodes that send something to the customer: if the message went out but the response was slow, you could send it twice.

Layer 2: decide what happens when a node fails

In the same node Settings you'll find On Error. It has three behaviours:

Stop Workflow: the workflow stops (the default). Fine for essential steps.
Continue: ignore the error and carry on. Use it only for optional steps, such as saving a copy to a log sheet.
Continue (using error output): the node gets a second output for errors. This is the most useful one: you can send the failed case down another path, for example replying to the customer with a fixed message ("I can't check that right now, someone from the team will get back to you shortly") and alerting your team.

A practical example in a WhatsApp AI agent: if the AI model node fails, the error output sends a polite reply and an alert to your phone. The customer isn't left hanging and you know what happened. It pairs nicely with handing off to a human.

Layer 3: an error workflow that always alerts you

The layers above work node by node. To leave no gaps, n8n lets you have an error workflow that runs whenever another workflow fails.

Step 1: create the error workflow

Create a new workflow and call it something like "Error alerts".
Add the Error Trigger node as its trigger.
After it, add the alert you prefer: Gmail (or your email), Telegram, Slack or a WhatsApp message to yourself.

In the alert, include what the Error Trigger gives you: the name of the workflow that failed, the error message, the last node that ran and the execution ID or link. That tells you exactly where to look.

A clear subject line helps a lot: "[n8n] Failed: WhatsApp agent – OpenAI node".

Step 2: connect it to your important workflows

In each workflow you want to watch, open Workflow Settings and under Error Workflow pick "Error alerts". Save. From then on, when that workflow fails in an automatic run, you'll get the alert.

You can use the same error workflow for all your workflows. One place for every alert.

Step 3: actually test it

Don't trust it until you've seen it work. An easy way: temporarily add a Stop And Error node (it throws an error on purpose with the message you write), let the workflow run automatically and check the alert arrives. Then remove that node.

Keep in mind that when you run a workflow by hand from the editor you see the error on screen; the error workflow is meant for automatic runs, the ones that happen when you're not watching.

Check your executions now and then

The Executions section in n8n shows every run, whether it succeeded and which node it stopped at. Five minutes a week looking at failures gives you clues: if a node keeps failing on odd data, you may need an IF node before it to filter those cases.

If you run n8n on your own server, remember to also keep an eye on the server itself. We cover that in the guide to installing n8n for free with Docker.

Quick checklist

Retry On Fail on nodes that call external services (AI, HTTP, Google).
On Error → Continue (using error output) on steps where a customer is waiting for a reply.
An "Error alerts" workflow with the Error Trigger and an email or Telegram alert.
That workflow selected under Error Workflow in every important workflow.
Test with Stop And Error, then remove it.
Five minutes a week reviewing Executions.

Start with templates that already handle failures

If you're just starting, first follow the guide to build your first AI agent in n8n. And to save hours, download the free WhatsApp agent or see the AgentKiosk pack with 8 agents ready to import.