What Is n8n? What It Does, What It Costs and Who It Fits
n8n connects your apps, APIs and AI models as nodes on a canvas and runs them on every trigger. What it is, what it is used for, what it costs and when it is the wrong tool.
n8n (pronounced “n eight n”) is a workflow automation platform. You connect apps, APIs, databases and AI models as boxes called nodes on a visual canvas, and n8n runs that chain every time a trigger fires: a form is submitted, an email arrives, a schedule ticks. You can use it as a hosted service (n8n Cloud) or install it free on your own server, and you can drop into JavaScript or Python wherever the ready-made nodes stop.
Businesses use n8n to move data between the tools they already pay for, to process documents and invoices, to send reports and alerts, and to run AI agents inside their operations. This page is the short version of each topic, with a link to the guide that goes deeper.
What does n8n stand for?
n8n is short for nodemation: “node” because the editor is a view of connected nodes and the software runs on Node.js, “mation” from automation. The founder found the full name too long to type in a terminal, so he shortened it the way developers shorten “internationalization” to i18n: keep the first and last letter and count the ones in between. n-odemati-o-n has eight letters between the two n’s, hence n8n. The story and the pronunciation are on n8n’s press page.
The company behind it, n8n GmbH, was founded in 2019 in Berlin by Jan Oberhauser, who is still its CEO. The same press page lists $240 million raised to date, a $180 million Series C led by Accel in October 2025, and a $5.2 billion valuation after a strategic investment by SAP in May 2026. For a buyer, the point of those numbers is simple: n8n is not a side project that might disappear next year.
How n8n works
Every n8n workflow is built from the same four parts:
- A trigger. The event that starts the workflow: a webhook from a form, a new row in a database, an incoming email, a schedule, a chat message, or a manual click while you test.
- Nodes. Each node does one thing: read from HubSpot, call an API, filter, branch on a condition, merge two lists, ask an AI model to classify a message, write to a spreadsheet, post to Slack.
- Items. Data moves between nodes as a list of JSON records called items. If a trigger returns 50 rows, the next node runs on all 50, one item at a time, unless you tell it otherwise.
- Executions. One run of the whole workflow, from trigger to last node, is an execution. n8n can save each one with the input and output of every node, so when something breaks you can open the run and see which node failed and on what data.
The canvas covers most work without code. When it does not, two escape hatches remain. Any field accepts expressions, small snippets that compute a value from earlier data. The Code node runs full JavaScript or Python on the items passing through. For an app without a ready-made node, the HTTP Request node calls any REST API directly; and when a team calls the same internal system from dozens of workflows, it can package that logic as its own node, which our guide to building n8n custom nodes walks through.

What is n8n used for?
n8n is used for any job where people copy data from one system to another, check it against rules, and tell someone what happened. In a company past its first million in revenue, that usually looks like this:
| Team | What n8n typically does | Systems it often connects |
|---|---|---|
| Sales and revops | Routes new leads, deduplicates and enriches them, creates deals, keeps two CRMs in sync | Web forms, HubSpot, Pipedrive, Salesforce, Slack |
| Finance | Creates invoices when a deal closes, reads supplier invoices, chases late payments, posts to the ledger | CRM, QuickBooks, Xero, Stripe, email |
| Operations and back office | Extracts data from documents, fills templates, moves files, runs approvals | Google Drive, SharePoint, PDF and OCR services, internal databases |
| Support | Triages tickets, drafts answers with an AI model, escalates what the model is unsure about | Zendesk, Intercom, help center, Slack |
| Management | Builds daily and weekly reports, alerts when a number moves out of range | Databases, spreadsheets, BI tools, email, Slack |
| Engineering and IT | Glues internal services together, provisions accounts, watches logs and releases | Internal APIs, GitHub, monitoring tools, identity providers |
Many of these workflows need a person to approve a step before it runs: a refund, a discount, an invoice above a limit. Chat is the usual place for that, and our guide to n8n Slack interactive buttons shows how to put Approve and Reject buttons in a Slack message and resume the workflow on the click.
n8n and AI agents
n8n is not an AI model. It is the plumbing around one: it fetches the data a model needs, sends the prompt, checks the answer, and writes the result where it belongs. It ships nodes for the major model providers, an AI Agent node that lets a model choose which tools to call, memory and vector store nodes for retrieval (RAG), and MCP client and server nodes, so an agent can use tools exposed over the Model Context Protocol and other AI clients can call your workflows as tools.
The useful pattern for a business is to mix both kinds of steps. Rules that must never bend (amounts, who gets emailed, what is legally required) stay in ordinary nodes. The model handles what rules cannot: reading a messy email, classifying a ticket, drafting a reply. A human approval step sits before anything that spends money or reaches a customer. The quality of the result still depends on the model, the data it sees and the checks around it.
n8n Cloud or self-hosted
The software is the same. What changes is who runs it.
| n8n Cloud | Self-hosted | |
|---|---|---|
| Who runs the servers | n8n | You, on your cloud account or your own hardware |
| Updates, backups, uptime | n8n | You |
| Where the data sits | n8n’s infrastructure | Wherever you put it |
| Software cost | Monthly plan | Free Community edition, or a paid Business or Enterprise license |
| Time to first workflow | Minutes | Quick for a test install; a production setup needs planning |
| Best when | No one on the team wants to run servers | Data must stay in-house, volumes are high, or you need custom nodes and packages |
Self-hosting gives you control over where the software runs. It does not, on its own, make you compliant: a workflow that sends customer data to an outside API still sends it. And “free” means no license fee; someone still has to patch, back up and watch the server. Our n8n installation guide covers the install options step by step, and self-hosted n8n set up and maintained for you is the service page if you would rather not run it yourself.
One point where n8n’s own sources disagree: its pricing page lists queue mode (the setup that spreads executions across several worker processes) as a Business plan feature, while its documentation lists queue mode as included without a paid plan and puts only multi-main mode behind one. If you plan to scale a self-hosted instance, confirm with n8n which applies to your edition before you size the project.
Is n8n free? What it costs
The self-hosted Community edition is free to use for your own business. n8n Cloud and the paid self-hosted editions are priced by executions, as listed on n8n’s pricing page (billed annually, checked November 2026; visitors in Europe see the same plans in euros):
| Plan | Price | Executions a month | Hosting |
|---|---|---|---|
| Community | Free | No plan limit | Self-hosted |
| Starter | $20 a month | 2,500 | Cloud |
| Pro | $50 a month | 10,000 | Cloud |
| Business | $800 a month | 40,000 | Self-hosted only |
| Enterprise | On request | Custom | Cloud or self-hosted |
The billing unit is what sets n8n apart: an execution is one run of the entire workflow, whether it has 3 nodes or 30 and whether it processes one record or a thousand. Only production runs count, the ones started by a trigger or a schedule; the runs you start by hand while building do not. Plans differ mainly in features such as SSO, separate environments, Git version control and shared projects. Our n8n pricing guide works through what each plan costs at real volumes and what self-hosting costs once you count the server and the hours.
Is n8n open source? The license in plain words
n8n’s code is public, but it is not open source under the usual definition. Most of it is under the Sustainable Use License, written by n8n, and n8n calls the model “fair-code”. The license text in n8n’s LICENSE.md lets you use and modify the software only for your own internal business purposes or for non-commercial or personal use, and lets you distribute it only free of charge for non-commercial purposes. Files marked “.ee” are not under that license at all: they need an n8n Enterprise license.
- Fine: running your company’s own workflows on a self-hosted instance, changing the code for your own use, hiring an agency to build and maintain workflows on your instance.
- Not fine without a deal with n8n: selling access to a hosted n8n to your customers, or making n8n part of a product you charge others for.
Is n8n easy to learn?
n8n is low-code rather than no-code. A first workflow, say a form that posts to Slack, can be built the same day. The learning curve starts when a workflow touches a real system: you read the app’s API documentation, set up credentials, understand how the data looks in JSON, and decide what happens when a step fails. People who already think in spreadsheets and have someone technical to review their work get there with practice. Teams with nobody comfortable around APIs usually start faster on Zapier or Make, or bring in help for the first builds.
What it takes to run n8n in production
Building a workflow that works once is the easy part. A workflow that runs every day for a year, on real data, while the apps around it change, needs a few things that most introductions skip:
- An error workflow. Each workflow can name an error workflow in its settings, which starts with the Error Trigger node and runs whenever an execution fails. Point it at a channel someone reads, with the workflow name, the failed node and a link to the execution.
- Retries and idempotency. APIs time out. Turn on retries on the nodes that call outside services, and design steps so that running them twice does not create two invoices or two deals.
- The right database. A self-hosted install uses SQLite by default. n8n also supports PostgreSQL, the usual choice for a production instance.
- Backups and execution data. Workflows and credentials live in that database: back it up, test a restore, and set how long execution history is kept so the database does not grow without limit.
- Updates on a schedule. n8n releases often. Pin a version, read the release notes, and test upgrades on a copy before production.
- Credentials owned by the company. Connect apps with service accounts the company controls, so nothing breaks when an employee leaves, and keep a list of what each credential can access.
- An owner. Every workflow needs a named person who gets the alert and decides what to do. A workflow nobody owns fails silently.
Our n8n best practices guide goes deeper into workflow structure, sub-workflows, naming, testing and monitoring.
When n8n is the wrong tool
n8n is a strong default for operations work, but not for every job:
| Your situation | Usually the better fit | Why |
|---|---|---|
| A few short automations, built by non-technical staff | Zapier or Make | More ready-made connectors, gentler learning curve, nothing to host |
| Everything already runs on Microsoft 365 and Dynamics | Power Automate | Deep native integration, often already in your license |
| Long, branching, high-volume or AI-heavy workflows; data that must stay in-house | n8n | Per-execution billing, code where needed, self-hosting |
| The process needs its own data model, screens and permissions for many users | Custom software, often with n8n beside it | A workflow tool is not a system of record |
If you are weighing n8n against the other two big platforms, the n8n vs Make vs Zapier comparison sets them side by side on price, hosting and fit. If you already run automations elsewhere, the guide on migrating from Zapier or Make to n8n covers what moves cleanly and what has to be rebuilt.
And if you would rather have the workflows built and kept running for you, 2V Automation is a certified n8n Expert Partner: the n8n automation agency page explains how a build works, on your own n8n instance, in your own accounts. The first step is the Automation Discovery Week ($2,000 for a team of 1 to 20 people): a week inside your operations that ends with a costed, sequenced plan of what to automate first.
Every guide on n8n in production
- n8n Slack Interactive Buttons: Approvals, Modals and File Uploads: Two ways to handle Slack buttons in n8n, Send and Wait approvals and a custom Block Kit app, with a full invoice upload example and fixes for clicks that fail.
- n8n Best Practices: How to Run Workflows in Production: What to set in n8n once workflows matter: error handling in three layers, silent failures, duplicates, secrets, execution data defaults, queue mode, and what n8n 3.0 changes.
- n8n Custom Nodes: How to Build, Deploy and Run Your Own: When a custom n8n node beats the HTTP Request node, how to build one with the official CLI, the three ways to deploy it, and what it takes to keep it running in production.
- n8n Pricing 2026: Every Plan, Self-Hosted Cost and Enterprise: Every n8n plan as priced on n8n.io in November 2026, what self-hosting and Enterprise really cost, and a worked example to pick the right plan. Updated November 2026.
- n8n Installation: Local in One Command, Production with Docker: Install n8n locally in one command, or set it up on a server with Docker Compose, Postgres and HTTPS, using the n8n 2.x settings most tutorials still get wrong.
Frequently asked questions
How do you pronounce n8n?
“n eight n”. n8n writes it in lowercase on its own press page, and the 8 stands for the eight letters between the two n’s of “nodemation”, the project’s original name.
Is n8n free?
The self-hosted Community edition is free: you pay for the server it runs on and the time to look after it. n8n Cloud is paid, from $20 a month for the Starter plan (billed annually, checked November 2026 on n8n's pricing page), after a free trial.
Is n8n open source?
Not in the strict sense. The code is public on GitHub, but most of it is under n8n’s Sustainable Use License, which limits use to your own internal business purposes or non-commercial use. n8n calls this model fair-code. For a company running its own workflows, the practical difference is small; for anyone who wants to resell n8n as a service, it is decisive.
Is n8n outdated now?
No. n8n ships releases continuously, added AI agent nodes, MCP support and human approval steps, raised a $180 million Series C in October 2025 and took a strategic investment from SAP in May 2026, according to its press page. Some older tutorials are outdated, because the editor and node names change often: check the date on any guide you follow.
Is n8n easy to learn?
Simple workflows can be built the same day if you are comfortable with spreadsheets and logins. Production workflows take longer, because you need to read API docs, understand JSON, and handle errors. Non-developers get productive faster with someone technical nearby to review their work.
Is n8n better than Zapier?
It depends on who builds and what runs. n8n is usually the better fit for long, high-volume or AI-heavy workflows, and when data must stay on your own servers. Zapier is usually faster for short app-to-app automations built by non-technical staff. Our comparison hub covers the trade-offs in detail.