What Is n8n? Nodes, Pricing and Who It Actually Suits (2026)

What is n8n, in one paragraph?

n8n is a tool for building automations that move data and fire actions between your apps. Its documentation home page, read 20 September 2026, puts it this way: "n8n is a fair-code licensed workflow automation tool that combines AI features with business process automation." In practice, it's the wiring between your form, your CRM, your inbox and an AI model.

The name throws people. The GitHub README, read 20 September 2026, has the answer: "It means 'nodemation' and is pronounced as n-eight-n." Founder Jan Oberhauser explains it in the same file. The "node-" part is "in the sense that it uses a Node-View and that it uses Node.js," and "-mation" is for automation.

Who it's built for is stated plainly. The homepage headline, read 20 September 2026, reads "AI agents and workflows for technical teams." Hold on to the word technical. It shapes almost every answer below.

How it actually works: nodes, workflows and triggers

Four words carry most of it: node, workflow, trigger and execution. The key concept glossary, read 20 September 2026, defines a node as "individual components that you compose to create workflows." A workflow is "a collection of nodes that automate a process." A trigger node is "a special node responsible for executing the workflow in response to certain conditions." An execution is one full run.

So a small automation might read like this. A form submission arrives, which is the trigger. Then one node looks the email up in your CRM, a second asks a model to draft a reply, and a third sends it. You wire them on the canvas, "the main interface for building workflows."

Two more pieces matter past the demo. Credentials "store authentication information to connect with specific apps and services," so you log into Gmail once and reuse it. Expressions "allow you to populate node parameters dynamically by executing JavaScript code." That second one is the fork in the road. The moment a workflow needs real logic, somebody is writing JavaScript.

Runs come in two kinds. The executions doc, read 20 September 2026, lists manual runs, which you start while testing. Production runs "run automatically" once you publish a version. That difference decides your bill.

Self-hosted or cloud: which are you signing up for?

n8n comes in two shapes, and people mix them up constantly. Cloud is hosted for you. Self-hosted means you install it on a server you own or rent. The deployment guide, read 20 September 2026, is blunt about the trade. For Cloud the guide lists installation as "No installation needed" and technical expertise as "None required." Self-hosted "Requires setup (npm, Docker, or server)."

The same page sorts buyers. "Want to start right away" points to Cloud. "Want to run n8n for free" points to self-hosted, where "You must provide and manage infrastructure" and maintenance is "Your responsibility."

One detail for anyone with a compliance question. Of hosted plans, the pricing page says "data is stored within the EU, on servers located in Frankfurt, Germany." If your data has to stay in the US, that's a self-hosting decision.

What does n8n cost, and what counts as an execution?

Three priced tiers are published on the pricing page, read 30 September 2026, plus Enterprise at "Contact Sales." Starter is 20 euro a month for 2,500 executions and Pro is 50 euro for 10,000, both hosted by n8n. Business is 667 euro for 40,000, and it is not a Cloud plan: the same page answers "Is the Business plan available on n8n Cloud?" with "Not yet. The Business plan is currently only available for self-hosted n8n." All three are billed annually, and every plan includes "unlimited users & workflows and every integration." The page prices in euros only, so confirm the dollar figure at checkout.

The execution is where people get surprised, in both directions. The pricing page argues its unit is the kinder one. "An execution is a single run of your entire workflow. It doesn't matter how many steps are in the workflow or how much data it processes." Add a fifth step to a four-step workflow and your bill doesn't move. Zapier would charge another task on every run.

The surprise runs the other way when a trigger fires more often than you expect. The executions doc spells out the rules. Schedule triggers "Count one execution every time the node fires, regardless of outcome." Webhook triggers count every inbound request that starts them, and that "includes requests with an empty body." A schedule set to every five minutes burns about 8,640 executions a month. Starter allows 2,500.

Some runs are free, which helps. Manual runs, sub-workflow runs, error workflow runs and "Polls that return no data" all "don't count towards your execution quota." Testing all afternoon costs you nothing.

Is it really free?

Yes, if you host it yourself. The edition comparison, read 20 September 2026, says you "can stick with the free Community edition indefinitely." It "includes almost the complete feature set of n8n." Cloud has no free tier. It offers a trial with "1,000 executions, 5 concurrent executions, a 180-second execution timeout," and "No credit card is required."

What the free edition leaves out is listed openly: custom variables, environments, external secrets, log streaming, projects, SSO and version control using Git. One omission matters more than the rest for a small team. Sharing is missing, so "Only the instance owner and the user who creates them can access workflows and credentials."

There's a free middle step too. Registering your Community instance with an email gets a license key that adds folders, debug in editor and custom execution data. You do it under Settings > Usage and plan, and the debugging feature alone is worth the two minutes.

Free, but not open source

This trips up buyers with a procurement policy. The license page, read 30 September 2026, says the Sustainable Use License lets you "use, modify, create derivative works, and redistribute, with three limitations." The first limits you to "your own internal business purposes or for non-commercial or personal use."

That page draws the conclusion itself. Citing the Open Source Initiative, it says "open source licenses can't include limitations on use, so we do not call ourselves open source." The model is called fair-code instead. The license FAQ draws the practical line. You may not "White-label n8n by removing our branding and offering n8n as is." You may "Charge your customers for workflow creation, setup, and maintenance services."

How does it compare with Zapier and Make?

Mostly the billing unit, and who is expected to build the thing. Zapier counts a task per successful action, Make counts a credit per module action, and n8n counts one execution per whole run. We priced six platforms side by side in our guide to AI automation tools for small business, so here we'll take the three people compare most.

Run one four-action workflow 500 times a month and the gap is plain. Zapier's pricing page says "A task is counted whenever Zapier successfully completes a unit of work for you," so call it 2,000 tasks. Make counts "one credit" per module action, so about 2,000 credits. The same work is 500 executions on n8n.

Price per unit runs the other way, so don't stop at the chart. Make lists 10,000 credits for $9 a month, which covers that workload five times over. The honest summary: n8n wins on long workflows and on control, Make wins on cheap volume, and Zapier wins on getting something live this afternoon. If your leads already sit in a CRM, building there first often beats all three, as our GoHighLevel and Zapier post argues.

Who it suits, and who should skip it

It suits you if someone on your side is comfortable in a terminal and happy writing the odd line of JavaScript. It doesn't suit you if nobody is. That isn't our opinion dressed up as a fact. The homepage sells "AI agents and workflows for technical teams," and the docs list "Don't have technical expertise" as a reason to buy Cloud.

Pick n8n when three things are true. You run high volumes of multi-step workflows, so per-execution billing saves real money. You have a named technical owner who patches the server and reads the logs. And you want your data on your own infrastructure.

Skip it when the opposite holds. A five-person service business wiring a booking form to a CRM will get there faster on a hosted tool. If the only possible maintainer is the owner, and the owner's evenings are full, self-hosted n8n becomes a liability the first time it stops. Our post on AI workflow automation covers the simpler versions of these jobs.

We build client automations mostly in GoHighLevel, with Claude on top. So treat this as an outside read of n8n rather than a war story. Everything above comes from n8n's own pages, read on 20 September 2026, with the quotes left intact so you can check them.

What breaks, according to the docs

The failure that costs the most is silence. A workflow can stop and nobody finds out, because n8n won't alert you unless you build the alert. The error handling doc, read 20 September 2026, has the fix: "For each workflow, you can set an error workflow in Workflow Settings. It runs if an execution fails." You build that second workflow yourself, starting with the Error Trigger node.

The same doc names the usual causes. Failures are "usually" down to "things like errors in node settings, or the workflow running out of memory." There's a gap in the alerting too. When the trigger itself fails, the report is thinner, because "the workflow doesn't execute."

Two more limits bite at volume. Cloud caps how many runs go at once, and the concurrency doc says anything past the limit will "queue for later processing." Starter allows five at a time. One line there matters mid-incident: "You can't retry queued executions."

The other limit is history. The pricing FAQ says n8n keeps saved executions until you hit a limit, "then remove the oldest." Starter's log retention is seven days. If a workflow has been quietly dropping one record a week, the evidence may be gone before you look. Build the error workflow on day one.

What is n8n worth trying for, and where should you start?

Start with the free trial rather than a server. The Cloud trial needs no credit card and gives you 1,000 executions, which is plenty to find out whether the canvas suits how you think. Build one workflow that matters, watch it for two weeks, then decide about self-hosting.

1. 01Count the runs first. Multiply each trigger by how often it fires in a month. A five-minute schedule is about 8,640 executions, which rules out Starter.
2. 02Build one workflow, not five. Pick the job that costs you most in hand-typing, then test it manually. Manual runs don't count against your quota.
3. 03Add the error workflow before you publish. One Error Trigger node into a Slack or email message. Name the person who reads it.
4. 04Only then decide on self-hosting. Plan on Docker, a server you patch, and backups you have actually restored once.

If the answer to "who fixes it when it breaks" is nobody, that isn't a reason to give up on automation. It's a reason to pick a tool your team can already support. The full six-tool comparison sits in our AI automation tools guide.

And if you'd rather not run any of this yourself, that's the work we do. We map the workflows, build them where your data already lives, and hand over error alerts with a named owner.