n8n can be run on your own server. That is one of the main reasons many companies and institutions choose it: your data stays under your control, and there is no fee per execution. But self-hosting also means taking responsibility for operations and security. This guide explains what is involved.
What you need
- A server, for example a cloud server from a German provider or a virtual machine in your own data centre.
- Docker as the environment in which n8n runs. This makes updates and migrations much easier.
- A database, usually PostgreSQL, for workflows, credentials and execution logs.
- A web server in front, which provides the encrypted connection, such as Caddy or nginx.
- Your own address, for example automation.your-company.com.
What it costs
The n8n Community Edition can be used without licence fees, but you should check the licence terms for your intended use. A suitable cloud server often costs only a few euros a month from German providers. The real effort lies in setup, securing the system and ongoing support. In return there are no costs per execution, which quickly makes a difference with many workflows. For a general overview, see the guide “What does automation cost?”.
Data protection
If n8n runs on a server in Germany, data leaves that server only where your workflows send it, for example to your CRM or an AI service. For the server itself, you conclude a data processing agreement with the provider. Document which workflows process which data. Your data protection officer will need that anyway.
What matters day to day
- Backups: Daily backups of the server and the database, tested regularly.
- Updates with care: n8n develops quickly. Updates occasionally bring changes that affect existing workflows. Back up before every update and check the most important workflows afterwards.
- Error notifications: A dedicated workflow reports when another one fails. Otherwise nobody notices.
- Securing access: Strong passwords, no open interfaces without protection, server access by key only.
- Keep an eye on storage: Execution logs keep growing. Old entries should be deleted automatically.
From practice
A typical stumbling block: after an update, Docker’s internal network configuration changed, and n8n could no longer reach the database. Cases like this are quickly fixed if you know where to look. Since then, checking exactly that point has been part of every update. More insights can be found in the “From the workshop” section on our AI page.
FAQ
How big does the server need to be?
For most SME use cases, a small cloud server with two processor cores and a few gigabytes of memory is enough. With many simultaneous workflows or large volumes of data, it is sized up accordingly.
Can we later move n8n to the cloud or the other way round?
Yes. Workflows can be exported and imported into another installation. Credentials have to be entered again.
