n8n is a tool that automates processes by linking your apps in workflows: a trigger (an email arrives, an invoice is created, it is Monday at 8:00) followed by a series of steps that read, transform and write data. It supports your own code and AI agents. You can use it in n8n's cloud or on your own server.
In practice
Typical uses in a small business: moving shop orders into the ERP, flagging when a customer has not paid, saving the attachments from an inbox into each client's folder, or sorting email with a language model.
Three things set it apart:
- Code when you need it. The Code node runs JavaScript or Python as just another step in the workflow.
- AI inside the workflow. You can connect models from OpenAI, Anthropic or Google, with tools and memory, and there is an AI agent node. You can require a person to approve the use of a tool before the agent uses it.
- You can install it yourself. The Community edition is free and can be hosted on your own infrastructure.
n8n charges per execution, and an execution is one complete run of the workflow, however many steps it has. As of October 2026, its cloud costs €20 a month on annual billing for the Starter plan (2,500 executions) and €50 for Pro (10,000), and stores data on servers in Frankfurt, according to its pricing page.
Its licence is not open source but fair-code: the Sustainable Use License lets you use it to run your business and charge your clients for building and maintaining workflows, but not offer it as a service for them to build their own. According to its FAQ, what counts is who controls the workflow logic.
The comparison with Make and Zapier is in what is n8n. We use it, among other tools, in our process automation projects.
Common mistakes
- Self-hosting with no maintenance plan. n8n's documentation leaves updates, backups and server security to you. Features such as SSO, environments or Git version control need a paid plan, even if you host it yourself.
- Assuming that hosting it in the EU makes the whole workflow European. If a step calls an AI model, that provider is another data processor, with its own processing regions.
- Credentials with more permissions than needed. Every workflow stores access to your apps. Give each one only what it needs.
- Letting a client edit workflows on your Community instance. At that point you are outside what the licence allows.