How n8n tells you about failures
n8n has a special workflow type called an error workflow. It starts with the Error Trigger node, and any other workflow can be pointed at it. When that other workflow fails, n8n starts the error workflow and passes it the details. The n8n error handling documentation covers the basics.
- A workflow failsan automatic run
- n8n starts your error workflowset under Settings
- Error Trigger passes the detailsname, error, last node, run link
- You get a messageSlack, email, Discord...
Set it up
- Create a new workflow called something like "Error alerts" and add an Error Trigger node first.
- Add a node that notifies you: Slack, Discord, email, Telegram, ntfy or whatever you actually read.
- Write a useful message. The Error Trigger gives you these fields, which you can drop into the message with expressions:
Workflow: {{ $json.workflow.name }} Error: {{ $json.execution.error.message }} Last node: {{ $json.execution.lastNodeExecuted }} Run: {{ $json.execution.url }} - Save it.
- Point each workflow at it. Open the workflow you want watched, open its settings and choose your workflow under Error workflow. Repeat for every workflow that matters.
Test it
The Error Trigger runs for failures in automatic executions. It does not run when you press the manual test button. To test, make a throwaway workflow with a Schedule trigger and a Stop and Error node, set its Error workflow to your alert workflow, activate it and wait for the schedule to fire.
Common gaps
- A workflow you forgot to point at the error workflow. Each workflow has its own setting. New workflows need it too.
- The error workflow itself failing. Keep it simple and test it again after you change it.
- Alert floods. A workflow that fails every minute sends a message every minute. Decide whether you want one alert per failure or a limit.
- Secrets in error text. Error messages can contain parts of a request or a URL. Think before forwarding them to a shared channel or an outside service.
- n8n itself being down. An error workflow cannot report that n8n stopped. Use a separate uptime check for that.
Polling instead of pushing
If you would rather watch from outside, n8n's public API can list failed runs. This needs the public API enabled and an API key:
curl -s -H "X-N8N-API-KEY: $KEY" \
"https://n8n.example.com/api/v1/executions?status=error&limit=20"
When an alert is not enough
An alert says something broke. You still have to read the error, work out why and decide whether to rerun. Pontavia's n8n Self-Heal with AI Triage ($49) is an n8n workflow that watches your own n8n for failed runs, checks simple rules first, and only asks Claude (Anthropic's AI) when the rules are not enough. You get one clear alert per failure: what broke, the likely cause, a suggested fix, and whether a rerun looks safe.
It flags credential problems straight away, retries only the workflows you list and only a limited number of times (nothing by default), and has a monthly spend cap on the AI step. Passwords, keys, tokens and email addresses are scrubbed before anything leaves your n8n. The AI only advises: nothing it says is ever run. It needs the n8n public API turned on and your own Anthropic API key, which you pay Anthropic for directly.
Frequently asked questions
Does the n8n Error Trigger work on manual executions?
No. It fires for failures in automatic (production) executions, not when you press the manual test button. Test it with a scheduled workflow that deliberately fails.
Can one error workflow handle many workflows?
Yes. Choose the same workflow under Settings, then Error workflow, in each workflow you want watched. The message can include the workflow name so you know which one failed.
How do I stop an n8n error workflow flooding me with alerts?
Decide on a limit up front, for example one alert per workflow per hour, or send a daily digest for noisy workflows. Also fix workflows that fail on every run rather than silencing them.
Does this work on n8n Cloud as well as self-hosted n8n?
Error workflows are a core n8n feature, so they work on both. The polling approach through the public API depends on your plan and settings, so check that the API is available to you.
Related guides
- Patch Proxmox Containers Safely with Snapshots
- Find Proxmox Guests With No Recent Backup
- Audit Proxmox Access: Users, API Tokens and 2FA
- All guides
Questions about your own setup? Send a message or book a free 15-minute call.