We operate n8n self-hosted on servers in German data centers: setup, data processing agreement, queue mode for load peaks, updates with test runs, daily backups and monitoring. If a workflow stops at night, you get a notification and we take care of it, instead of someone noticing it by chance on Monday.
The Docker instance is up after an afternoon, and that's enough for a test too. As soon as real invoices, customer data and deadlines run through the workflows, questions arise that don't come up in the tutorial.
This page covers operations, not building the workflows. What n8n is and what can be automated with it is on our general n8n page.
We set up n8n as a Docker deployment on a server in a German data center, with its own domain, TLS certificate, reverse proxy and separate database. You get the data processing agreement, the list of subprocessors and documentation of which data is located where. This way you have an answer instead of a shrug at your next audit.
Impact Audit-ready instance instead of shadow ITAs soon as several workflows start simultaneously, standard operation is no longer enough. We switch n8n to queue mode with Redis and separate worker processes, so executions are processed in parallel and a long-running workflow doesn't block the rest. The number of workers is based on your actual volume, not on a sales pitch.
Impact Parallel runs instead of a queueWe roll out new n8n versions first on a test instance and run your critical workflows against them. Only once they pass does the version move to production, in a time window you know. If a problem does occur, we roll back to the previous version, because the old image and the database backup are still there.
Savings 2–4 h per version changeWe monitor not just whether the server responds, but whether the workflows actually complete. Failed executions trigger a notification by email or Teams, with the workflow name and error message. We additionally set up error workflows so that a single API error doesn't silently halt an entire route.
Impact Errors noticed in minutes instead of daysWorkflows, credentials and execution data are backed up daily, we test the recovery and don't just write it down. Access runs through named users instead of a shared admin password, credentials are stored encrypted in n8n and not as plain text in the workflow. If an employee leaves the company, you revoke access and the rest keeps running.
Impact Tested recovery instead of hopeNot every operation has to self-host. We tell you honestly when your own server pays off and when the cloud variant is the more sensible choice. What matters is the data protection requirement, volume and the question of who is on call in an emergency.
Self-hostedThe instance runs on German servers under your hosting contract or ours, the data doesn't leave the tenant. Sensible as soon as personal data, customer documents or systems from your own network are involved. You can read what n8n can do in general on the n8n service page.
You can hand operations entirely to us or run it yourself and only bring us in for updates, incidents and expansion. Either way, you get access to the server and workflows, there is no black box and no lock-in. You retain control and can bring operations back in-house at any time.
If your day-to-day already runs entirely in Microsoft 365, Power Automate is often the shorter path and the license is usually already paid for. n8n plays to its strengths when systems outside the Microsoft world, custom code or strict data sovereignty come into play. Customers often end up running both side by side.
From the initial assessment to a monitored production instance in 30 days, at a fixed price agreed in advance instead of an open-ended timesheet.
In 60 minutes we clarify volume, data protection requirements and the current state of any instance that may already be running. After that you know whether self-hosting is even the right path for you.
We set out server sizing, queue mode, backup schedule, monitoring and responsibilities in writing. You get a fixed price for setup and a clear monthly amount for ongoing operations.
We set up the instance, move over existing workflows and test them with your real data. If an instance is already running, we migrate it with an announced cutover window instead of during live operation.
You receive access, documentation and the data processing agreement. After that we take over updates, monitoring and incidents, or we show your team how to do it themselves. You retain control over the server and workflows.
This page answers the operational questions: server, data protection, updates, monitoring. If you're more interested in which processes can be mapped with n8n and how a workflow is built, start on our general n8n page.
In the free initial analysis, we look at your instance or your project for 60 minutes and tell you honestly whether self-hosting in Germany makes sense. No obligation and no sales pressure.
Self-hosted n8n on a server in Germany is a good foundation for GDPR-compliant operation, but the technology alone is not enough. You additionally need a data processing agreement with the hoster and with us as the operator, a permissions concept, deletion periods for execution data and documentation of the data flows. That is exactly what we set up. As soon as a workflow calls an AI model outside the EU, that one step must be assessed separately, the rest of the instance remains unaffected.
The costs are made up of the server, the operation and the changes, and we give them to you after the initial analysis as a fixed price instead of a vague range. The server share is the smallest item for most mid-sized companies, because n8n runs on modest hardware. It only gets expensive when no one is responsible for updates and incidents and an outage hits operations.
You need queue mode as soon as several workflows run at the same time or individual executions take a long time, for example with document processing or AI calls. In standard operation, n8n processes executions in the same process, and a long-running workflow then slows down everything else. In queue mode, Redis distributes the executions across several workers that can scale independently of one another.
n8n releases new versions at short intervals, and anyone who doesn't update for months accumulates security gaps and an uncomfortably large version jump. We roll out updates on a fixed schedule, always first on a test instance with your critical workflows, then in production in the announced time window. If you want to run operations yourself, we show your team this procedure and remain reachable in the background.
That belongs in the contract before the first workflow goes live, and we write it in there too. If we take over operations, we agree on response times and availability, and monitoring reports the outage to us, not first to you. If you run operations yourself, the responsibility lies with you, and we tell you that honestly beforehand instead of hiding it in the fine print. For critical routes, we additionally build it so that incoming data is not lost in an outage but is caught up afterwards.
