I Replaced a Bunch of Cron Scripts with n8n (And Sleep Better)

A year ago my “automation stack” was a folder of Node scripts, a forgotten crontab, and prayers. Whenever something broke, I’d ssh in, read logs, and swear quietly.

n8n is the first tool in a long time that made me retire scripts without feeling like I’d given up control. It’s open source, you can self-host, and — crucially — you can see the data at each step instead of sprinkling console.log everywhere.

The one-sentence pitch

Visual pipelines: trigger → do thing → branch if error → Slack me so I don’t find out from a user.

If you’ve used Zapier or Make, the mental model is the same. If you haven’t: imagine Lego, but each brick is “HTTP request” or “Postgres query.”

Concrete example: this blog’s pipeline

Roughly what my publish flow does (yours doesn’t need to match):

Step	What actually runs
1	WhatsApp Business webhook receives my message
2	Code node strips my shorthand (tags: dev, rust) into fields
3	OpenAI node returns Markdown with YAML frontmatter
4	GitHub creates/commits src/content/blog/{slug}.md
5	WhatsApp replies with “live at …” so I know it worked

The part that used to hurt in bash? Step 4 — parsing messy model output, escaping quotes, not committing garbage. In n8n I can pin the failing execution, copy the payload, and fix the prompt without redeploying a server.

Example: the “parse message” shape

I’m not sharing my full prompt (it changes weekly), but the data object my code node expects looks roughly like this:

{
  "raw": "topic: n8n tips\n tone: casual\n tags: automation, n8n",
  "topic": "n8n tips",
  "tags": ["automation", "n8n"]
}

If topic is empty, I short-circuit and send myself an angry Telegram message instead of committing nonsense. Learned that the hard way.

How I debug when step 4 looks “fine”

1. Open the execution, expand every node output — not the summary.
2. Copy JSON into a scratch file and diff against the last good run.
3. If GitHub returned 422, read the body — GitHub’s error messages are annoyingly precise once you scroll past the HTML panic.

Half my “n8n is broken” moments were expired credentials or a branch protection rule I forgot existed.

Credentials: treat them like milk

Rotate API keys on a calendar reminder. Name them after the workflow (n8n-blog-publish-prod) so when you see one in the audit log you know what to revoke. I used to have three keys all labeled “automation” and it was a security and sanity problem.

Why not plain Node?

I could ship a single Express app. For a long-running product I might still. But for glue code, n8n wins on:

• Retries and error branches without writing the same boilerplate again
• Secrets UI instead of .env drift across three machines
• Non-engineer readability when I show someone else what broke

Try it in ten minutes

Cloud is fastest if you just want to play:

1. Sign up at n8n.io — free tier exists
2. Build “Webhook → Set fields → HTTP Request → Slack”
3. Break it on purpose and watch the execution log

Self-hosting one-liner (good for homelab people):

docker run -it --rm -p 5678:5678 n8nio/n8n

Then open http://localhost:5678 and click around.

When I wouldn’t use n8n

• Tight latency sub-50ms paths where you need compiled code and profiling.
• Heavy transformation that’s really a product (ETL that should live in dbt or a data warehouse).
• Secrets you can’t trust a third party with — self-host fixes part of that; air-gapped doesn’t.

Use the right hammer. n8n is glue, not a database.

---

Automation pays off in the boring moments: when a pipeline fails loudly at step three instead of silently for a week. n8n made those failures visible enough that I actually fix them.