The short version
n8n Cloud and self-hosted n8n run the same workflow engine. The difference is who operates it and where your data lives.
n8n Cloud is the right default for most small teams. You get a working instance in minutes and n8n handles upgrades, uptime and platform security. Self-hosting makes sense when data residency, volume, internal network access or custom nodes outweigh the cost of running it yourself.
Where n8n Cloud wins
No operations work. No servers, no database backups, no upgrade weekends. For a company without an engineer who enjoys infrastructure, that is worth a lot.
Fast start. You can build and test workflows the same day, then revisit hosting once you know your real volume.
Predictable responsibility. When the platform has a problem, n8n fixes it. With self-hosting, your on-call person does.
Verified nodes only. The restriction to verified community nodes is also a safety feature. Unvetted packages are a supply-chain risk on any self-hosted instance.
Where self-hosted n8n wins
Data control. Every payload, credential and execution log stays in your AWS, Azure or Google Cloud account, or your own data center. For regulated industries this often decides the question, and it simplifies conversations with auditors.
Cost at volume. The Community Edition has no execution fees. A team running hundreds of thousands of executions a month pays for servers instead of allowances.
Internal network access. A self-hosted instance can sit next to an internal Postgres database or a private API without exposing either to the internet. That matters when connecting an LLM to internal data.
Full configuration. You control environment variables, execution timeouts, data retention, task runners and scaling. Queue mode with Redis and worker processes lets one instance handle much larger loads.
Custom and community nodes. Unverified community nodes and your own custom nodes are only available when you self-host.
What self-hosting actually involves
A production self-hosted deployment needs Docker or Kubernetes, Postgres, automated backups, TLS, an auth layer in front of the editor, uptime monitoring and alerting on failed executions, and a routine for upgrades. Budget a few hours a month for maintenance on a small instance, more for a queue-mode cluster. If that work will not get done, choose Cloud.
What we recommend
Start on n8n Cloud unless you already know data residency or volume rules it out. Move to self-hosted when a specific requirement appears, not on principle.
Lumeor sets up and maintains n8n on either option. Our n8n automations work shows the kind of workflows we build, and the cost of an n8n automation program explains how we price builds and ongoing care.