n8n is a workflow automation platform: it watches for a trigger, then runs a series of steps — fetch, transform, decide, send — without you touching anything. It’s open source, you can run it on your own server for a flat cost, and it has first-class AI integrations, which is why it has become the orchestration layer of choice for people building AI-powered businesses.
We run our entire content and business infrastructure on a self-hosted n8n instance — well over a hundred production workflows as of this writing. This tutorial is the on-ramp we’d hand a beginner: what the pieces are, your first real workflow, and the habits that separate demos from systems.
The five concepts that make everything else make sense
- Workflows. A workflow is a canvas of connected nodes. Data enters at a trigger and flows left to right through each node.
- Triggers. Every workflow starts with one: a schedule (“every morning at 7”), a webhook (something calls a URL), or an app event (“new row in Airtable”). No trigger, no workflow.
- Nodes. Each node does one job — fetch an RSS feed, call an AI model, filter items, send an email. n8n ships with hundreds; the HTTP Request node covers anything without a dedicated node.
- Items. Data moves between nodes as a list of items (JSON objects). Most beginner confusion is item confusion: a node runs once per item unless told otherwise. Internalize this early and n8n becomes predictable.
- Credentials. API keys and logins are stored once, encrypted, and referenced by nodes — never pasted into workflows.
Cloud or self-hosted?
n8n offers a hosted cloud (paid plans, zero server work) and a self-hosted community edition (free software; you pay for a small server). Our recommendation is unambiguous for anyone planning real volume: start on cloud or a local install to learn, and move to a self-hosted VPS the moment a workflow matters — flat infrastructure cost beats per-execution anxiety, and your workflows are assets that live on hardware you control. The full walkthrough is in our self-hosting guide, and the platform trade-offs are in n8n vs Zapier vs Make.
Your first real workflow: a daily niche digest
Tutorials that end at “Hello World” teach you nothing about real automation. This one is small but genuinely useful: every morning, fetch the latest posts from feeds in your niche, format a clean digest, and email it to yourself. It exercises a trigger, data flow, transformation, and delivery — the skeleton of almost every workflow you’ll ever build.
Shortcut: download our importable starter workflow — n8n-starter-workflow.json — then in n8n: Workflows → Import from File, and follow along to understand each node.
Step 1: Schedule Trigger
Add a Schedule Trigger node, set it to run daily at your chosen hour. This is the heartbeat.
Step 2: RSS Read
Add an RSS Read node and point it at a feed in your niche (most blogs and news sites expose one — try adding /feed to a WordPress site’s URL). Execute the node and inspect the output: you’ll see items, each with title, link, and snippet. This inspection habit — run a node, read its output — is 80% of learning n8n.
Step 3: Limit and sort
Add a Sort node (by date, newest first) and a Limit node (keep 5–10). Real feeds are noisy; workflows curate.
Step 4: Format the digest
Add a Code node that assembles the items into one tidy email body — a loop over items building an HTML list. The starter workflow includes this code with comments. (This is also where you’d later add an AI node to summarize each item — one node, and your digest becomes an AI product.)
Step 5: Send it
Add an email node (SMTP or Gmail), create the credential when prompted, wire it up, and execute the workflow end to end. Then toggle the workflow to Active. You now have software working for you on a schedule — that feeling is the whole point.
From first workflow to production habits
The gap between a demo and a system is a short list of habits. Adopt them from workflow one and you’ll skip months of pain:
- Name nodes for what they do. “Fetch niche RSS,” not “RSS Read1.” Six months from now, you are the stranger reading this workflow.
- Add an error path. n8n supports error workflows — a catcher that notifies you when something fails. A silent failure in an active workflow is the most expensive kind.
- Keep state outside the workflow. Anything that must survive between runs (queues, dedupe lists, statuses) belongs in a store like Airtable, a database, or a sheet — not in your head or the workflow itself.
- One workflow, one job. When a canvas needs scrolling in two directions, split it. Chained small workflows beat one monolith — we learned this the hard way refactoring a 44-node monster into four clean stages.
- Export your workflows. They’re JSON files. Version them, back them up — they’re assets. Some people sell them.
Where AI fits
n8n treats AI models as nodes: OpenAI, Anthropic, and any HTTP-reachable model slot directly into the flow, and there are agent and vector-store nodes for more advanced patterns. The practical pattern that scales: deterministic nodes for everything that shouldn’t vary, model calls only where judgment is needed — and route cheap models to mechanical steps, reserving strong models for the few steps that earn them. That single discipline is the difference between an AI workflow that costs cents and one that surprises you at month’s end. Deeper patterns live in the AI Agents hub.
What to build next
- A content repurposer — new blog post triggers social-post drafts into a review queue
- A lead responder — form submission triggers an enrichment lookup and a personalized reply draft
- A monitoring agent — daily check of prices/competitors/keywords with a summary to your inbox
- A publishing pipeline — queue in Airtable, content generated and formatted, posted on schedule
Each of these is a beginner-plus build using exactly the skeleton you just learned: trigger → fetch → transform → deliver.
Frequently asked questions
Do I need to code to use n8n?
Mostly no. The node editor is visual, and expressions cover most data-wrangling. A little JavaScript in Code nodes unlocks a lot, and AI assistants write those snippets well — treat code as an occasional power tool, not an entry requirement.
Is the free self-hosted version really enough?
For a solo builder, comfortably. The community edition runs unlimited workflows and executions on your own server; paid tiers add team and enterprise features. Check n8n’s current licensing page when you set up, as tiers evolve.
How is this different from Zapier?
Pricing model and depth. Zapier charges per task, which gets painful at pipeline volume; n8n self-hosted is flat-cost. Zapier is easier for the first hour; n8n is stronger by the tenth workflow. Full breakdown: n8n vs Zapier vs Make.
What server do I need to self-host?
A small VPS runs n8n comfortably for solo workloads — our step-by-step VPS guide covers sizing, Docker setup, HTTPS, and backups.
Next: Self-host n8n on a VPS · n8n vs Zapier vs Make · or the full Automation hub.