What n8n actually does.
n8n (pronounced "n-eight-n", short for nodemation) is a workflow automation platform built by n8n GmbH, a company founded in Berlin in 2019. The idea fits in one sentence: an event in one tool triggers a chain of actions in other tools, with no human in the loop.
A form is submitted: the contact is created in the CRM, enriched, qualified, and your sales rep gets a summary in Slack. An invoice reaches its due date: the reminder goes out on time, and stops as soon as the payment lands. Each step is a node you connect to the others in a visual editor.
Three things set it apart from similar tools:
- Code when you need it. A node does not do exactly what you want? Write a few lines of JavaScript or Python right inside the workflow.
- Your choice of hosting. In n8n's cloud, or on your own server, where the data stays with you.
- A massive community. n8n's GitHub repository has 205,891 stars (as of September 27, 2026), one of the most followed automation projects in the world.
What you can automate with n8n.
n8n is not just about connecting two apps. In SMBs, these are the uses that give back the most time:
| Process | What runs on its own | What changes |
|---|---|---|
| Invoicing and reminders | Scheduled reminders, stop on payment, reconciliation | Cash collected faster, nothing forgotten |
| Sales | Lead enriched, scored, created in the CRM with a summary | Sales calls back knowing who they are talking to |
| Reporting | Exports pulled, numbers computed, dashboard up to date | No re-typing, decisions on fresh numbers |
| Documents | Email attachments detected, filed, archived | A clean folder for accounting, no manual sorting |
| Support | Simple requests handled, the rest escalated with context | Answers even at night, the team on what matters |
A full real-world sequence, with reminder templates, is in Automated Invoice Reminders. Eight workflows described node by node are in n8n Workflow Examples.
Cloud or self-hosted: the two versions of n8n.
n8n comes in two versions sharing the same engine. Here are the plans n8n lists in euros as of September 27, 2026, billed annually (17% cheaper than monthly):
| Plan | Price | Executions / month | Hosting |
|---|---|---|---|
| Community | Free | No license cap | Your server |
| Starter | €20/month | 2,500 | n8n cloud |
| Pro | €50/month | 10,000 | n8n cloud |
| Business | €667/month | 40,000 | Your server only |
| Enterprise | Custom quote | Custom | Cloud or your server |
Every plan includes unlimited users, workflows and integrations. In the cloud, n8n bills per workflow execution, whatever its complexity: a 30-step workflow counts as one execution. That is the major difference from tools that bill every step.
When self-hosted, the Community edition covers almost everything. The paid-only features mostly matter to large teams: SSO, environments, shared projects, Git version control, workflow and credential sharing.
The full math, with the hidden costs of each option, is in n8n Pricing 2026. Setting it up on your own server is covered in Self-Hosted n8n.
n8n and AI: agents, tools and MCP.
This is where n8n has pulled furthest ahead. It connects to the main model providers (OpenAI, Anthropic, Google and others) and lets you combine several in one workflow. Above all, it can build real AI agents:
- The AI Agent node takes a request, reasons, and picks the tools to use on its own: search the CRM, send an email, update a sheet.
- Memory lets it hold a conversation or remember the context of a case.
- Human approval: for sensitive actions (sending a message, changing or deleting data), the workflow pauses and asks for your approval in Slack, Teams, Telegram or n8n's chat. You approve, the action runs. You deny, nothing happens.
- MCP (Model Context Protocol) both ways: an n8n agent can call external MCP servers, and an n8n workflow can itself become an MCP server. Since version 2.13, you can even create and edit workflows by describing them to Claude or ChatGPT.
The step-by-step method to build one is in How to Build an AI Agent with n8n. To understand MCP, see MCP (Model Context Protocol).
Install n8n in a few minutes.
To try n8n on your computer, Docker is enough. The two commands from the official documentation, with a timezone set:
docker volume create n8n_data
docker run -it --rm --name n8n -p 5678:5678 \
-e GENERIC_TIMEZONE="Europe/Paris" -e TZ="Europe/Paris" \
-e N8N_ENFORCE_SETTINGS_FILE_PERMISSIONS=true \
-v n8n_data:/home/node/.n8n \
n8nio/n8n
n8n is then available at http://localhost:5678. The n8n_data volume keeps your workflows across restarts. Replace the timezone with yours.
For production, the documentation now recommends Docker Compose rather than this single command, with a PostgreSQL database, a domain name and HTTPS. It also states that self-hosting is meant for experienced users: a misconfiguration can cost you data or expose your instance. Without server skills in-house, the cloud or a partner who operates the instance for you are the sensible options.
n8n vs Make and Zapier.
| Criterion | n8n | Make | Zapier |
|---|---|---|---|
| What is billed | The workflow execution, however long | Each module action (one credit) | Each successful action (one task) |
| Self-hosting | Yes, free Community edition | No | No |
| Learning curve (our view) | Medium | Easy | Easiest |
Sources: n8n, Make and Zapier pricing pages, checked on September 27, 2026. In practice, a 10-step workflow run 1,000 times counts as 1,000 executions in n8n, and up to 10,000 billed units in Make or Zapier.
Our rule of thumb: Zapier to start very fast on two or three simple automations, n8n as soon as volume, logic or AI matter. Make sits in between. The details, including when we recommend something other than n8n, are in n8n vs Make vs Zapier.
When not to choose n8n.
n8n is not the answer to everything, and saying so saves time:
- You have two simple automations and no technical profile. Zapier or Make will be up and running faster, at a modest cost at that volume.
- You want self-hosting with nobody to run it. Updates, backups, security: a server needs looking after. Without in-house resources, pick the cloud or have it operated for you.
- You need SSO, separate environments or Git version control. These require a paid plan, Business when self-hosted, or Enterprise.
In every other case, and especially once AI enters your processes, n8n is the tool we deploy by default. If you want us to set it up and run it for you, that is what we do: our services.
n8n: what people ask us.
Is n8n free?
Yes when self-hosted: the Community edition is free and covers almost every feature. You only pay for the server, often €5 to €20/month for an SMB. In the cloud, plans start at €20/month (billed annually) for 2,500 executions. Details in our n8n pricing guide.
Is n8n open source?
Not strictly. The code is public, but n8n is distributed as fair-code under the Sustainable Use License: internal use in your company is free, commercial use included. What is restricted is reselling n8n itself as a service.
Do you need to code to use n8n?
Not for everyday automations: you connect blocks in a visual editor. You can go further with JavaScript and Python inside workflows. Self-hosting, however, requires real server skills, and n8n recommends it for experienced users.
n8n or Zapier: which one should I choose?
Zapier is still the fastest for two or three simple automations. n8n pulls ahead as soon as volumes grow, logic gets complex or AI comes in, because it bills per workflow execution rather than per step. The full comparison is in n8n vs Make vs Zapier.
Where is the data stored with n8n?
In the cloud, on n8n's infrastructure. When self-hosted, on your server: you choose the host and the country. That is often what convinces companies handling customer or financial data.
Which AI models can you use in n8n?
n8n connects to the main providers, including OpenAI, Anthropic and Google, and can combine several models in the same workflow. You pay for model usage directly to the provider.