n8n Tutorial: A Step-by-Step Learning Path (2026)

This page is the starting point for learning n8n in the RoamerHost Guides. It puts our whole n8n series in the order you should read it, so you can go from zero to workflows that actually run in production. You start with how n8n works, build your first workflow, connect webhooks, Google Sheets and LINE, make workflows run on their own schedule and alert you when they fail, and finish with AI Agents and RAG knowledge bases. You only need two things to begin: a Google account and an n8n instance with an HTTPS address (self-hosted, n8n Cloud or managed hosting all work, and we compare them below). At 30 minutes a day, the first five stages take about two weeks. Each stage explains what the articles teach and when to read them, so skip whatever you already know.

Facts checked as of October 2026. Everything here follows n8n 2.x; at the end of October the stable release is the 2.45 line. Button and node names match the n8n editor.

Before you start: accounts and an n8n with HTTPS

Many people start by running n8n in Docker on their own laptop at localhost:5678. That is perfectly fine for the first two stages, because dragging nodes, inspecting data and writing expressions do not need a public address. The trouble starts in stage three. An outside service can only send data to n8n if it can reach it, which means a public HTTPS URL. LINE's Messaging API is explicit about this: the webhook URL must use HTTPS with a certificate issued by an authority that mainstream browsers trust, and self-signed certificates are rejected. Google OAuth makes an exception for localhost, but once your n8n moves to a real domain, the redirect URI has to be HTTPS. Rather than moving halfway through, decide where n8n will live before you begin.

As for accounts, besides the Google account you will use for Google Sheets, set up one channel for alerts (Telegram, Gmail or Slack) for the error-handling stage. To send LINE messages you need a LINE Official Account with the Messaging API enabled in its admin console. For the AI stage you will also need an API key from a model provider such as OpenAI, Google Gemini or Anthropic Claude. Model usage is billed by that provider, no matter where your n8n runs.

Self-hosting

The n8n community edition is free, and getting it running in Docker is easy. Keeping it reachable and stable over the long run is the harder part: you need a domain, a reverse proxy and an SSL certificate, plus a few environment variables. If you leave GENERIC_TIMEZONE unset, it defaults to America/New_York and your schedules run on New York time. Behind a reverse proxy, set N8N_WEBHOOK_URL to your public address (it replaced the older WEBHOOK_URL in 2.35.0); otherwise the editor shows webhook URLs with the internal port 5678, and nothing outside can reach them. Also note that n8n 3.0, whose first beta shipped on October 27, no longer supports installing with npm or npx, so if you self-host today, start with Docker or Docker Compose and you will not have to migrate later. You also have to follow releases and security advisories yourself. October 2026 alone had two advisories that required upgrades; see n8n October security updates. This route suits people who know Linux and Docker and will set aside time for maintenance.

n8n Cloud

The official cloud service works as soon as you sign up, with the URL and HTTPS already in place, so it is the quickest way to get going. Plans are tiered by monthly workflow executions. When we checked, the official pricing page listed Starter at 20 euros a month billed yearly for 2,500 executions, and Pro at 50 euros a month billed yearly for 10,000 executions. That is plenty while you learn. Once workflows go live and schedules run more often, you will need to watch the execution count, and community nodes are more restricted on Cloud. The trade-offs are covered in n8n self-hosted vs n8n Cloud.

Managed hosting

The third option is a managed host. Someone else takes care of the server, SSL, updates and backups, and you get your own n8n instance with a fixed HTTPS address, usually without per-execution billing. It suits people who want to focus on building workflows and plan to put them to real use right after learning. RoamerHost offers this, and we cover it at the end of the page.

If you are still deciding, read the real cost of self-hosting n8n, which also puts a price on your setup and maintenance time. If you are choosing between n8n, Zapier and Make, read n8n vs Zapier vs Make.

The learning path

The path has seven stages. The first five are the fundamentals everyone needs, and when you finish them you will have workflows that run on their own and tell you when something breaks. Stage six covers AI, and stage seven covers long-term operations and platform choices, which you can read when they become relevant. The times below are rough estimates for working through each stage the first time, including some troubleshooting.

The first five stages add up to roughly 7 to 8 hours, which works out to about 30 minutes a day for two weeks. The order is deliberate. Expressions come after your first workflow because the idea of data flowing through as separate items only clicks once you have built something. Error handling comes after scheduling because you only need alerts once workflows run while you are not watching. AI Agents come last because the tools an agent uses are the same nodes and workflows you learned to build earlier.

Stage 1: understand what n8n does

What is n8n explains the basics in plain language. Every workflow starts with a trigger, such as a form submission, a webhook call or a scheduled time. Data then travels along the connections from node to node, and each node does one job. The article also covers the kinds of services n8n connects to, real examples, and the differences between self-hosting, n8n Cloud and managed hosting. Read it if you have never used n8n or are still deciding whether it fits your work. If you already know you want n8n and just want to build, go straight to stage two. If you are coming from Zapier, also read migrating from Zapier to n8n, which explains why copying your Zaps one by one leaves you with many small workflows that are hard to maintain.

Stage 2: build a workflow, then understand how data moves

Build your first workflow walks you through a three-node workflow: a Manual Trigger, a Google Sheets node using Append Row, and a Slack or Telegram message. Along the way you learn the canvas, nodes and credentials, how to test with Execute workflow, how to read each node's OUTPUT panel, and finally how to switch to an automatic trigger and click Publish. Step one uses the RoamerHost dashboard to log in; if you self-host or use n8n Cloud, log in to your editor and start at step two. If the Google credential step stalls, jump to the credentials section of the Google Sheets guide, then come back.

The n8n expressions guide is the article most worth your time in the whole series. Data passes between nodes as an array, each entry is an item, and most nodes run once per item, so three orders reaching a messaging node produce three messages. When a result surprises you, the cause is usually an item count that differs from what you expected. The guide starts there, then shows how to read values from the previous node, how to reach further back with $('Node Name'), how to format dates with $now, and when to use Edit Fields, If, Switch, Merge, Split Out and Aggregate, before discussing when the Code node is worth it. Read it right after your first workflow, or the first time a field preview shows [undefined]. Two patterns you will use constantly:

Hi {{ $json.customer }}, your order {{ $json.order_id }} has shipped
{{ $now.toFormat('yyyy/MM/dd HH:mm') }}

The first line drops fields from the current item into text. The second prints the current time in your n8n time zone, for example 2026/10/27 08:00.

Stage 3: connect outside services

This is where you need the HTTPS address mentioned earlier. Read the three guides in order (webhooks, Google Sheets, then LINE), because each one builds on ideas from the one before.

n8n webhooks and third-party integrations is about letting other systems notify you. Put a Webhook node at the start of a workflow and n8n gives you a URL. Paste it into a form tool, Stripe, GitHub or LINE, and whenever an event happens the other side sends the data over and your workflow starts immediately. You will learn how to set the HTTP method and path, how to read incoming data with {{ $json.body.email }}, how to keep strangers out with Header Auth or a hard-to-guess path, and how to manage credentials. Read it whenever you want things handled the moment they happen, instead of polling for new data every few minutes. While learning, you can call the test URL yourself with curl:

curl -X POST "https://n8n.example.com/webhook-test/new-order" \
  -H "Content-Type: application/json" \
  -d '{"name":"Alex Chen","email":"[email protected]"}'

Click Listen for test event on the Webhook node first, then send the request, and the node output will show name and email under body.

The n8n Google Sheets guide is usually the first project people put to real use. The slowest part is the Google Cloud credential setup. First you choose between OAuth and a service account. With OAuth, you copy the OAuth Redirect URL from the n8n credential screen into Google Cloud exactly as shown, because a single wrong character produces redirect_uri_mismatch. Then you cover the Append Row, Update Row, Get Row(s) and Append or Update operations, plus two complete examples: writing each form submission to a new row, and reading the sheet every morning to list items due today. Read it if your data already lives in a spreadsheet or if stage two stalled on Google credentials. It also covers tokens that expire after 7 days and running out of API quota.

The n8n LINE messaging guide handles one of the most common requests in Taiwan: sending notifications to LINE. LINE Notify, which many older tutorials rely on, shut down on March 31, 2025, and n8n's built-in Line node, which depended on it, has been deprecated since 1.64.0, so those tutorials no longer work. The current approach uses the Messaging API of a LINE Official Account: create the account, enable the Messaging API in its admin console, get a channel access token, and call LINE's push message API from an HTTP Request node. The guide also shows how to capture your own userId or a group ID through a webhook, and explains that push messages count toward the account's monthly message allowance, which is 200 messages a month on the free plan in Taiwan and usually enough for system alerts sent to yourself. Read it if you or your customers should receive alerts on LINE, or if you want your daily reminders delivered there.

Stage 4: make workflows run on time

The n8n schedule guide shows how to use the Schedule Trigger for fixed times such as every 15 minutes, 8 a.m. daily, 9 a.m. on weekdays or the first of each month. It covers what to fill in for each Trigger Interval option, how to write cron expressions (weekdays at 9 a.m. is 0 9 * * 1-5), and how to work around requests cron cannot express, like the last day of every month. The most important part is the section on time zones: a self-hosted n8n without a time zone setting runs on New York time, so an 8 a.m. schedule fires in the evening in Taiwan. Anything that happens daily, weekly or monthly belongs on a schedule. If the other system can notify you instead, a webhook from stage three is faster and avoids piles of empty executions.

Stage 5: know when something breaks

Once workflows run on their own overnight, the real risk is a failure nobody notices for days. The n8n error handling guide splits the job into two layers. At the node level, Retry On Fail retries flaky API calls, and On Error decides whether a bad item stops the run or gets routed around. At the workflow level, an Error Trigger workflow sends you a Telegram message or email whenever any workflow fails, including a link to the execution so you can retry from the failed node. It also shows how to use Stop and Error to fail a run on purpose when nothing errored but the result is wrong, and how to deal with APIs that return 429 rate-limit responses. Set this up the same week your first real workflow goes live, then attach every new workflow to the same error workflow.

Stage 6 (advanced): AI Agents and RAG knowledge bases

Everything up to now follows a fixed "when A happens, do B" pattern that you drew in advance. The n8n AI Agent guide takes a different approach: give the AI a goal and some tools, and let it decide whether to look something up and which tool to use first. You open a test chat window with the Chat Trigger, attach a Chat Model, Memory and Tools under the AI Agent node, learn to write clear tool descriptions, and use the System Message to say what the agent should do when it cannot find an answer. You will also see how to estimate cost and why Max Iterations should not be raised. It reads best after the first five stages, because the tools an agent uses are the HTTP Request, Google Sheets and sub-workflow skills you already have.

If you want the AI to answer from your own documents, such as product manuals, internal policies or FAQs, continue with building a RAG knowledge base with n8n and Qdrant. It uses two workflows. The ingestion workflow splits documents into chunks, creates embeddings and stores them in Qdrant. The Q&A workflow gives an AI Agent Qdrant as a tool, so it retrieves relevant passages before answering. The most common mistake is using different embedding models for ingestion and search, which makes results poor no matter what else you tune.

By this point you will want to combine n8n with other tools. Dify plus n8n explains a split where Dify understands requests and n8n carries out the actions, which suits assistants that need to both understand people and get things done. n8n plus NocoDB covers moving workflow data into a proper database once Google Sheets starts to struggle.

Stage 7: operations, security and cost

You do not need these on day one, but you will after a few months of running workflows. n8n execution data growth and cleanup explains why the editor slows down over time. n8n stores every execution along with each node's input and output, and a frequent schedule can add thousands or tens of thousands of records a month. The fix is a retention period and not saving successful executions while still keeping failed ones. Read it when the editor feels sluggish or the executions list takes a long time to load.

If you self-host, treat n8n security updates as required reading. It covers what the two October 2026 advisories fixed, the minimum version for each release line, how to back up before upgrading in Docker, and why you must keep N8N_ENCRYPTION_KEY safe, since without it none of your credentials can be decrypted. If your n8n exposes webhooks to the internet, follow the official advisories regularly.

If you are deciding where to run n8n, read three articles together. Self-hosted vs n8n Cloud compares execution-based billing, where your data goes and community nodes. The self-hosting cost breakdown converts setup, upgrades and backups into money to show when self-hosting pays off. n8n vs Zapier vs Make looks at USD subscriptions, per-task pricing and data leaving the country from a Taiwanese team's point of view. Read them before you pay for anything, or when your company asks you for a purchasing evaluation.

A two-week plan at 30 minutes a day

Here is one way to fit the first five stages into 14 days. If a day runs long, push the rest back. What matters is that you build every step in your own n8n, because reading without doing will catch up with you in stage three.

Week one covers the fundamentals. On day 1, read What is n8n, get your instance ready, log in and create the owner account. On days 2 and 3, follow Build your first workflow to make the three-node workflow (manual trigger, write to a sheet, send a message). Before the end of day 3, swap in a Schedule Trigger and click Publish, then check the next morning that it really ran on its own. On days 4 and 5, read the expressions guide and practice on that same workflow: put the name and a formatted date into the message, feed in three items on purpose to watch it send three messages, then use Aggregate to combine them into one. On days 6 and 7, move to webhooks. Click Listen for test event, call the test URL with the curl command above and study the data, then publish, switch to the production URL and find that run in the Executions tab.

Week two connects real services. On days 8 and 9, build both examples from the Google Sheets guide: form submissions written to a row, and a morning summary of items due today. On day 10, follow the LINE guide to create an Official Account, enable the Messaging API and send the due-date reminder to LINE instead. On day 11, read the schedule guide, open the workflow settings to confirm the Timezone is Asia/Taipei, and use cron to move the reminder to 8 a.m. on weekdays. On day 12, read the error handling guide, build an Error Trigger workflow, set it as the Error Workflow in the settings of every other workflow, then break a node's URL on purpose and confirm the alert reaches you. On days 13 and 14, tidy up: give each workflow a clear name, add a Sticky Note describing what it does, and export each one as JSON for safekeeping. If you have time left, start the AI Agent guide.

After two weeks you will have four or five workflows running for real, and you will hear about it when one breaks. From week three on, pick from stages six and seven as you need them.

Three things beginners get stuck on

Mixing up the Test URL and the Production URL

The Webhook node has two URLs. The Test URL uses the /webhook-test/ path and only listens after you click Listen for test event (or run the whole workflow). The official docs say it stays active for 120 seconds, and incoming data shows up right in the editor so you can build as you go. The Production URL uses the /webhook/ path, only works after you publish the workflow, and its data does not appear on the canvas; you find those runs in the workflow's Executions tab. The two classic mistakes are pasting the Test URL into a form tool and forgetting to change it at launch, so every submission gets a 404, and switching to the Production URL but staring at the canvas waiting for data. If you self-host behind a reverse proxy, also check that the editor shows your real domain in the webhook URL. If it does not, N8N_WEBHOOK_URL is probably missing.

Forgetting to publish, or not publishing again after changes

n8n 2.x replaced the Active toggle from 1.x with Publish. A clean run with Execute workflow only means your manual test passed. Schedules, production webhook URLs and forms only start working after you click Publish at the top of the canvas. Version 2.x also adds an idea that 1.x did not have: your edits are saved automatically as new versions, but production keeps running the version you last published, so if you change a workflow and do not publish again, the old version is still the one running. In 1.x, saving an active workflow changed production immediately, which is why people coming from older versions trip over this. If you only change workflow settings, such as the time zone, n8n republishes automatically.

Leaving the time zone unset

A self-hosted n8n without GENERIC_TIMEZONE defaults to America/New_York. An 8 a.m. daily schedule then runs at 8 p.m. Taiwan time, or 9 p.m. after US daylight saving time ends. The fix is to add GENERIC_TIMEZONE=Asia/Taipei and TZ=Asia/Taipei to the container and recreate it. To run a single workflow in another zone, set Timezone in that workflow's settings. One related trap: JavaScript's native new Date() inside a Code node ignores n8n's time zone setting, so use $now and $today for dates instead. The schedule guide covers all of this in its time zone section.

What changed between n8n 2.x and older 1.x tutorials

n8n 2.0 arrived in December 2025, and many tutorials and videos online still show the 1.x interface. Most of the concepts still apply, but the clicks will not always match. A quick way to tell whether a tutorial is old: if there is an Active toggle in the top right of the editor, it is 1.x.

The first change is the one above, the Active toggle becoming Publish. When an old tutorial says to turn Active on, click Publish in 2.x; when it says to turn Active off, choose Unpublish from the dropdown next to Publish. The second is renamed nodes. The Set node in older tutorials now appears as Edit Fields (Set), and searching for "set" still finds it. Window Buffer Memory, common in AI tutorials, is now called Simple Memory. The third is the Item Lists node, which disappeared from the node panel in 1.21.0 and was split into six standalone nodes: Split Out, Aggregate, Sort, Limit, Remove Duplicates and Summarize. Where an old tutorial uses Item Lists to split an array, use Split Out.

A few other 2.0 changes stop older instructions from working. The Start node is no longer supported, so use a Manual Trigger for manual runs. The Code node blocks access to environment variables by default, so the old trick of reading secrets with $env in a Code node gets blocked. The Execute Command and Local File Trigger nodes are disabled by default. And any tutorial that sends notifications with LINE Notify or the old Line node no longer works, whatever version it was written for.

The next major version is already on its way. n8n 3.0 released its first beta on October 27, 2026, while the stable release is still the 2.45 line. According to the official 3.0 breaking changes, the legacy Function, Function Item and Item Lists nodes are removed, so workflows that still use them will fail, and npm installs are no longer supported. For learning, stick with the stable 2.x line. If you self-host, open the Migration Report under Settings before moving to 3.0 to see which workflows are affected.

If you would rather not run the server yourself

Once you finish this path, you may notice that most of the time goes into the server side rather than the workflows themselves: HTTPS certificates, time zones, version updates, security advisories, backups and execution cleanup. If you would rather just build workflows, RoamerHost's fully managed n8n hosting runs in an isolated container with GENERIC_TIMEZONE=Asia/Taipei already set, includes a dedicated HTTPS subdomain with fixed webhook URLs, places no limits on executions or workflows, handles updates and backups automatically, offers support in Chinese, and costs NT$5,988 billed yearly (avg. NT$499/mo). Everything in this series works the same on self-hosted n8n, n8n Cloud or managed hosting, so pick the option you can comfortably maintain.

FAQ

Q: Can I learn n8n without knowing how to code?

Yes. None of the first five stages require programming. Dragging nodes, picking fields and writing simple expressions is enough. Expressions look like code, but most of the time they are just lookups like {{ $json.fieldName }}, and a few rounds of practice with the expressions guide will make them familiar. You rarely need the Code node, and you can learn it when you do.

Q: Can I learn on my own computer first?

Yes. Stages one and two work fine on localhost, and Google OAuth accepts localhost redirect URIs. You get stuck when outside services need to reach you: LINE webhooks require HTTPS with a trusted certificate, and form tools and Stripe also need to reach your instance. Schedules also stop whenever your computer is off. Before stage three, move n8n somewhere with a fixed HTTPS address.

Q: Are older 1.x tutorials still useful?

Mostly, yes. Triggers, nodes, items and expression syntax have not changed much. When following along, translate a few things: the Active toggle is now Publish, Set is now Edit Fields (Set), Item Lists is now separate nodes such as Split Out and Aggregate, and the Start node is replaced by Manual Trigger. Tutorials built on LINE Notify cannot be followed at all; use the Messaging API approach instead.

Q: How long until I have a workflow running in production?

On the plan above, your first self-running workflow is live on day 3. After two weeks you will have several workflows that use webhooks, Google Sheets and LINE, run on a schedule and send error alerts. Allow about one more week for AI Agents and RAG. Your real pace depends on having an actual task you want to automate; practicing on a real need is much faster than copying examples.

Sources

• n8n docs: Save and publish workflows, 2.0 breaking changes, 3.0 breaking changes, Deprecated nodes
• n8n docs: Webhook test and production URLs, Configuring N8N_WEBHOOK_URL, Time zone variables
• n8n docs: Edit Fields (Set), Line node deprecation; n8n blog: Introducing n8n 2.0; n8n releases on GitHub
• n8n: n8n Cloud pricing (checked October 2026)
• LINE Developers: Build a bot and webhook URL requirements, LINE Notify end-of-life notice; Google: OAuth redirect URI rules