n8n

At n8n, a workflow with twenty steps costs the same as a workflow with two. The platform charges per execution: one complete run of a workflow, however many steps it contains. That one difference largely decides who will find n8n cheaper than the competition.

n8n was founded in Berlin in 2019 by Jan Oberhauser. In October 2025 the company raised $ 180 million in a Series C, at a valuation of $ 2.5 billion. Since then the focus has been on AI: n8n presents itself as the place where you build AI agents and regular automation together.

The license is important to understand. n8n calls its model fair-code and uses the Sustainable Use License. The source code is public on GitHub and you may use and modify the software for free. That applies to internal use in your company, non-commercial use and personal use. You may not sell or host n8n as a service for others. So formally it is not open source by the strict definition, even if in practice it looks like it.

With many automation tools you pay per step or per task. A workflow that reads a form, updates a CRM and sends a message then counts three times. At n8n the same workflow counts once.

It gets more generous still. Only production executions count. Manual test runs, sub-workflows, error workflows and checks that find no new data don't count, according to the pricing page. Webhook requests that n8n rejects aren't counted either.

For simple integrations that makes little difference. For long workflows with lots of branches it can make a big one. Work out in advance how many complete runs you expect per month. That number decides which plan you need, not the number of steps.

You can use n8n in two ways. In the cloud, everything runs at n8n itself, with the data in Frankfurt. Or you host it yourself, for example with Docker on your own server.

The self-hosted Community Edition is free, with unlimited executions. It includes nearly all features: the editor, all integrations, code steps in JavaScript and Python, and queue mode for scaling. A free license key adds folders and debugging in the editor.

What's missing is collaboration. In the Community Edition, only the owner of the instance and the creator can see a workflow and its credentials. Shared projects, SSO, environments and Git version control are in the paid plans.

Also good to know: n8n collects telemetry by default, but you can turn it off. Paid self-hosted licenses report the number of executions to the n8n license server daily.

n8n charges a flat fee per plan, not per user. In the paid plans, users and workflows are unlimited. Prices are in euros, and paying yearly saves 17%.

Starter costs € 24 per month, or € 20 per month billed yearly (€ 240 per year). For that you get 2,500 executions per month and 5 concurrent executions. Pro costs € 60 per month, or € 50 per month billed yearly, for 10,000 executions and 20 concurrent. There is also a larger Pro tier at € 145 per month for 50,000 executions.

Business costs € 800 per month, or € 667 per month billed yearly (€ 8,000 per year), for 40,000 executions. That plan is for self-hosting only and includes SSO, environments and Git version control. Enterprise is custom, in the cloud or self-hosted.

The pricing page also lists a free cloud plan with 50 executions per month for 1 user. Enough to try it out, not for real work. You can try the paid cloud plans for 14 days, with 1,000 executions and no credit card.

Two pieces of fine print. The AI credits in the cloud plans can't be topped up and don't roll over to the next month. And n8n doesn't give prorated refunds when you cancel; a yearly subscription runs until the end of the term.

What made n8n big in recent years is AI. With the AI Agent node you build agents that call tools themselves, with multiple agents or with your own documents as a source. You can connect any model you want, including local models that run offline. If you don't want to send sensitive data to a cloud model, that helps a lot.

There are also features to keep AI in check. A human can approve a step before the agent continues. With evaluations you test how reliable an AI workflow is. Through MCP you build workflows from other AI apps. The built-in n8n Assistant is still in preview.

Zapier runs only in the cloud and charges per task. Choose Zapier if you want to connect lots of business apps quickly without thinking about servers or code. For long workflows with many steps, n8n can work out cheaper because it charges per complete run.

n8nFrom € 20/moFree planZapierFrom $ 19.99/moFree plan

Make is also visual, but only works in the cloud. Activepieces is open source and also self-hostable, with a simpler setup. n8n is the logical choice if you have technical skills in-house, want to use code in your workflows or build AI agents with your own models.

The self-hosted Community Edition is free, with unlimited executions. In the cloud there is a free plan with 50 executions per month, and a 14-day trial with 1,000 executions.

An execution is one complete run of a workflow, regardless of the number of steps. Manual tests, sub-workflows and error workflows don't count.

You can use n8n for your own business processes. The Sustainable Use License doesn't allow you to resell n8n as a service or host it for others.

In the cloud, the data stays within the EU, on servers in Frankfurt. If you self-host, it lives wherever you run your server.