What is n8n? How it works, what it costs, and who it fits

What n8n is

n8n is a workflow automation tool. A workflow is a chain of steps, called nodes, that you connect on screen. The first node, the trigger, waits for an event: a form submission, a set time, an incoming message. The nodes after it read, transform and send data from one app to another.

An example: each time an order comes in on your online store, the workflow adds the customer to a spreadsheet, pings the team in chat and drafts the invoice.

The vendor is n8n GmbH, a German company founded in 2019. The tool comes in two forms, and that is the first thing to understand before you look at prices:

• n8n Cloud: the vendor hosts the tool and you pay a subscription based on a monthly quota of executions;
• the Community edition: you install n8n on your own server. The vendor charges nothing, but the server, the updates and the backups are on you.

The guide n8n Cloud vs self-hosted compares the two setups point by point.

What n8n does in practice

On the number of integrations, the vendor’s pages contradict each other: "500+ native integrations" (home page); "2,000+ integrations" (integrations page, community nodes included) source. The second figure also counts nodes built by the community and by partners. Custom nodes are available only when you self-host.

The number of workflows is not capped: Unlimited active workflows on every plan. The limits of the Cloud plans sit elsewhere:

• run time: up to 5 minutes per execution on Starter, 40 minutes on Pro;
• concurrency: 5 executions at once on Starter, 20 on Pro at the 10,000-execution level. Beyond that, executions wait in a queue and run in order of arrival;
• history: logs kept for 7 days on Starter, 30 days on Pro.

For support, the pricing grid says: Forum (Starter, Pro, Business); dedicated support with SLA (Enterprise).

What n8n bills

You can work out what that means. Take two workflows:

• an eight-node workflow started by each order, 800 orders a month: 800 executions. The node count changes nothing; eight steps or two count the same;
• a two-node workflow scheduled every 15 minutes: 4 runs an hour × 24 hours × 30 days = 2,880 executions a month, whether or not there is anything to process.

The small scheduled workflow uses more than three times as many executions as the big one. On its own, it goes past the 2,500 executions of the Starter plan.

Executions you start by hand from the editor, sub-workflows and error workflows do not count toward the quota, so building and debugging a workflow does not eat into it.

The plans in brief

Two plans are hosted by n8n. Starter costs €24 per month (€20 per month billed annually) for 2,500 executions. Pro starts at €60 per month for 10,000 executions, with a second level at 50,000. The prices recorded are in euros.

Business, from €800 per month, is a paid plan you install on your own server, for companies with fewer than 100 employees that want SSO, environments or Git version control. Enterprise has custom pricing.

The full table and the volume levels are further down, with the date recorded. The n8n pricing page works through three scenarios in executions, and the cost calculator compares one scenario across several tools.

Is n8n free?

Yes if you self-host, no on Cloud.

• On n8n Cloud, there is no permanent free plan. The free trial lasts 14 days, comes with the features of the Pro plan, is limited to 1,000 executions and asks for no credit card. If you do not move to a paid plan, the instance is deleted; your workflows stay downloadable for 90 days.
• The Community edition is free and needs no license key. It does not include everything: custom variables, environments, projects, SSO, sharing of workflows and credentials, and Git version control are left out.

Free does not mean open source. The code is published under this license: Sustainable Use License 1.0 ("fair-code"); ".ee." files under the n8n Enterprise License source. It allows use for your own internal business purposes. It forbids hosting n8n as a service where your customers build their workflows, or offering it as a white-label product. The guide Is n8n free and open source? covers what is allowed.

Hosting and data

On where data is stored, the pricing page says: Hosted plans: EU, servers in Frankfurt (Germany); self-hosted: wherever n8n is installed source. For a US business, that means the hosted plans keep your data in Germany; if it has to stay in a particular place, self-hosting is the setup that lets you choose.

When you self-host, the docs recommend installing with Docker. The Docker Compose guide asks for at least 4 GB of RAM and 2 vCPUs, for a stack that includes the Assistant sandbox. The default database is SQLite; PostgreSQL is supported. The vendor reserves self-hosting for experienced users: the guide how to self-host n8n follows the steps in the documentation.

One thing to know: telemetry, meaning usage data sent to the vendor, is on by default. You can turn it off when you self-host, but not on Cloud, where sessions in the interface are recorded and deleted after 21 days.

AI and agents

An AI agent is a program that lets a language model decide which actions to take. For n8n, the pricing page says: AI Agent node included (Cloud and Community); models paid with your own keys or Gateway credits. MCP, the protocol that lets an AI assistant call outside tools, is covered too: MCP Server Trigger and MCP Client included in every Cloud plan and in the Community edition source.

Two counters sit next to the execution quota:

• model calls: you pay for them with your own provider keys, or with Gateway credits, a prepaid balance offered on Cloud Starter and Pro only;
• Assistant credits, for help building workflows: 1,600 a month on Starter, 4,000 or 9,600 on Pro depending on the level, reset on the first day of the month and not carried over. Once they run out, the Assistant pauses; your workflows keep running.

Which setup for your situation

• You want to start without installing anything: the n8n Cloud free trial, then Starter if your workflows fire fewer than 2,500 times a month.
• You want to pay nothing for the software and you can run a server (or someone does it for you): the Community edition.
• Your workflows run every few minutes: count the runs first. A schedule every 5 minutes comes to 12 × 24 × 30 = 8,640 executions a month, which calls for the Pro plan.
• You want to sell automation to clients: read the license first. It forbids hosting n8n as a service where your customers build their own workflows.

Alternatives to n8n

The difference that weighs most is the billing unit. Make counts one credit per module run, the trigger included: a long scenario uses more than a short one, while n8n ignores length. Zapier counts tasks, meaning successful action steps, and does not count triggers. Activepieces, like n8n, counts one credit per run and offers a free Community edition you host yourself.

The head-to-heads n8n vs Make and n8n vs Zapier run the math on the same scenario. For a wider view, see the n8n alternatives.

What the documentation does not say

For the Business plan, the vendor’s pages contradict each other: the grid shows executions per month, while the FAQ on the same page refers to an annual quota for annual subscribers. Prices are recorded in euros, as shown in the plan table; dollar prices are not recorded.