n8n: A Complete Guide for Beginners

Q: Do I need to know how to program to use n8n?

No. Basic workflows are assembled with a mouse, and {{ }} expressions can be picked up in an evening. Understanding JSON structure helps, and the Code node with JavaScript is an option for advanced users, not a requirement.

In short

n8n is an open-source automation platform: services are wired together with visual nodes into workflow chains, and for complex logic you drop an AI agent inside. The main difference from Zapier and Make: n8n can be deployed on your own server for free — no execution limit and full control over your data. The cloud version starts at $20/month (as of July 2026), but it's paid with a foreign card, so in Russia people more often choose self-hosting via Docker. The learning curve is gentle: a "form submission → Telegram notification" workflow takes 15–20 minutes to assemble with a mouse. And you don't have to assemble it by hand at all: an n8n workflow is JSON, and Claude Code can write it from a text description.

What n8n is and why everyone is talking about it

n8n (pronounced "n-eight-n", short for nodemation) is an open-source automation builder that a Berlin team has been developing since 2019. Three concepts are enough to get started:

A workflow is a chain of steps that runs by itself: "request arrives → validate → send to Telegram → write to a spreadsheet".
A node is one step in that chain: receive a webhook, call an API, send a message. There are over a thousand ready-made integration nodes.
A trigger is the starting node: the event a workflow begins with (a form, a webhook, a schedule, a new email).

Why n8n is on everyone's lips right now: the platform embedded LLMs directly into the builder earlier than its competitors and became a popular assembly ground for no-code AI agents. The scale is confirmed by money: in October 2025 the company raised $180 million at a $2.5 billion valuation, and in May 2026 SAP took a stake — already at $5.2 billion. According to n8n, more than 1.5 million people use the platform every month.

One honest caveat: the n8n license is fair-code (Sustainable Use License). For yourself and for a business's internal tasks — free rein; reselling n8n itself as your own cloud service is not allowed.

Self-host or Cloud: what should a beginner pick?

There are two options: n8n Cloud (sign up and go) and self-hosting, where the Community Edition is free. Prices as of July 2026:

Option	Price	Executions/month	Who it suits
Self-host (Community)	0, you only pay for your VPS	unlimited	anyone who can run Docker; your data stays with you
Cloud Starter	$20–24/month	2,500	trying it out without a server
Cloud Pro	$50–60/month	10,000	a business with dozens of scenarios
Business	from $800/month	40,000	companies: SSO, Git versioning, isolated environments

The key billing detail: one execution is one run of the entire workflow, no matter how many steps it contains. Zapier charges per operation, which makes long scenarios in n8n several times cheaper.

For Russia the picture is simple: Cloud requires a foreign card, self-hosting doesn't. It deploys with a single command:

docker run -it --rm -p 5678:5678 -v n8n_data:/home/node/.n8n docker.n8n.io/n8nio/n8n

After it starts, the interface opens at localhost:5678. For production use a VPS with 1–2 GB of RAM is enough — n8n is undemanding. A condensed take on n8n with a ready recipe for "payment → grant access → receipt → CRM" is in the arsenal card.

How the interface is arranged

Everything happens on the canvas: node search on the left, the workflow itself in the middle. Four things worth knowing before your first run:

Credentials — a separate vault for keys and tokens. A service key is added once and reused across all workflows; it's never exposed inside the nodes. If API keys are new territory for you, here's a primer on keys and secrets.
Expressions — data from previous nodes is substituted with constructs like {{ $json.name }}; the easiest way is to drag fields in with the mouse from the left panel.
Execute Workflow — the test-run button: you see the input and output of every node, and debugging becomes very visual.
Executions — a log of every run with its data and errors. The first place to look when "nothing works".

Your first workflow step by step: form → Telegram

A classic small-business task: a form submission should land in Telegram instantly. No code required:

Create a bot through @BotFather and save the token. Send the bot /start — otherwise it won't be able to send you a single message (the most common reason for "the bot is silent").
Find your chat_id — through @userinfobot, for example.
In n8n, create a workflow and add the n8n Form Trigger: n8n has built-in forms, so no separate service is needed. Define the fields — name, phone, question. If the form already lives on your site, use the Webhook trigger instead and paste its URL into the form settings.
Add a Telegram node with the Send Message operation. Paste the bot token into Credentials and your chat_id into the Chat ID field.
Build the message text with expressions: New request: {{ $json.name }}, {{ $json.phone }}.
Click Execute Workflow, open the test form link, submit a request — the message should arrive in Telegram.
Flip the Active toggle. Important: a form and a webhook each have two URLs — the test one works only while the editor is open, the production one works after activation. Mixing them up is the second most popular beginner mistake.

Fifteen to twenty minutes and you have a working automation you wouldn't be embarrassed to show a client.

How do you build an AI agent in n8n?

AI Agent is a node with three components plugged into it: a chat model, memory and tools. The difference from a rigid chain: the agent decides for itself which tool to call and in what order, based on the task.

Models: OpenAI, Anthropic (Claude), Google, Mistral, and local ones via Ollama. You'll need the provider's API key — for Claude from Russia that's a solvable problem, covered here.
Tools: HTTP requests, Google Sheets, another workflow used as a tool — and MCP servers, which n8n connects to directly. What MCP is — in this guide.
Memory: Simple Memory is fine for testing but only lives until a restart; for production connect Postgres or Redis.

A typical agent: reads a customer question from Telegram, looks up the answer in a knowledge base, replies on its own, and escalates hard cases to a human. More working scenarios with an assessment of the payoff are in the article on AI agents for business.

A trick: let Claude Code write the workflow for you

An n8n workflow is an ordinary JSON file. Which means a coding agent can write it. There are two working approaches:

Without integration. Ask Claude Code or Cursor: "generate an n8n workflow JSON: a form with fields X, validation Y, a Telegram notification". Paste the resulting JSON straight onto the n8n canvas (Ctrl+V) or use Import from File. One nuance: the agent can't fill in credentials — after importing, open every service node and select your own Credentials, otherwise the workflow won't run. That's the cause of most "the import doesn't work" complaints.
Through MCP. The open-source n8n-mcp server connects an agent to your n8n over the API (the key is created in Settings → API). Then Claude Code creates workflows in your instance itself, runs them, reads errors from Executions and fixes them — a loop with no manual copy-pasting.

The quality of the result depends not on n8n but on how you state the task — how to phrase it so the agent gets it on the first try. If Claude Code is still unfamiliar, start with the beginner's guide. And to make the agent build workflows with discipline — decomposing the task, checking itself, not inventing nodes that don't exist — that discipline comes from the Quest engine: ready-made rules and roles for Claude Code and Cursor that turn "build me an automation" into a manageable process.

FAQ

Do I need to know how to program to use n8n?

No. Basic workflows are assembled with a mouse, and {{ }} expressions can be picked up in an evening. Understanding JSON structure helps, and the Code node with JavaScript is an option for advanced users, not a requirement.

Is n8n really free?

The Community Edition is: no limit on workflows or executions, you only pay for the server. The fair-code license has a single restriction: you can't resell n8n itself as your own cloud service. Automating your own business and building projects for clients is fine.

What makes n8n better than Zapier or Make?

Self-hosting with control over your data, billing per workflow execution (not per operation), and native AI agents. Zapier is simpler in the first half hour, and Albato has more connectors to Russian services — n8n wins on complex, high-volume scenarios.

Does n8n work in Russia?

Self-hosted — with no restrictions: it's just software on your server. n8n Cloud is paid with a foreign card. AI nodes need a model provider's key — the ways to get a Claude key from Russia are described in the article linked above.

What kind of server do I need for self-hosting?

To start, a VPS with 1–2 GB of RAM and Docker installed is enough. The load grows with the number of active workflows and the volume of data, but a typical small business lives on a hosting provider's minimum plan for years.