The n8n production checklist nobody gives you

A working workflow isn't a production setup. The gap between "it runs on my machine" and "I can run this for a paying client and sleep at night" is a handful of unglamorous things nobody puts in the tutorials. Here's the checklist.

1. Backups you don't have to remember

n8n doesn't back itself up, and manual JSON exports don't happen consistently. Automate it — pull workflows via the API and commit them somewhere you own. The one time you need a backup is the time you forgot to take it. (More: n8n backup.)

2. Version history + diffs

Saving overwrites the previous version, so "what changed last week?" is unanswerable by default. You want a snapshot per change and a readable diff — not raw JSON. (More: n8n version control.)

3. Secrets out of your workflow JSON

Workflows hide hardcoded credentials in headers and parameters. Before you back up, commit, or share one, strip them. See exactly what leaks with the free redaction demo.

4. Monitoring & alerts

n8n won't tell you when a workflow is edited, deactivated, starts failing, or the instance goes offline. You need external monitoring that pushes to where you actually look — Slack, Telegram, Discord. (More: n8n monitoring.)

5. A rollback plan

When something breaks, "rebuild it from memory" is not a plan. Keep known-good versions you can restore as importable JSON in one step. (More: n8n disaster recovery.)

6. An audit trail (especially for clients)

When a client asks "what changed?", you want a real answer — a timestamped history of every change per workflow, per client. (More: n8n audit log.)

The short version

• Automated, redacted backups ✔
• Version history with real diffs ✔
• Secrets stripped before anything leaves ✔
• Alerts on change / failure / offline ✔
• One-click rollback ✔
• Per-client audit trail ✔

You can wire these up yourself, or get all six from one agent that runs next to your n8n — Keel, free on one instance. Either way: don't run client automations without them.