Once you decide n8n is the right automation tool, the next question is whether to run it on n8n cloud or self-host. This sounds like a deployment detail; it is actually a decision about how you want to operate automation as your workflows grow.
When n8n cloud is the right call
You should use n8n cloud when:
- You are getting started and have no DevOps capacity.
- Your workflows are <50 and execution volume is moderate.
- You have no data residency or compliance requirements.
- You value zero ops over predictable cost at scale.
For most teams shipping their first 5-10 automations, cloud is the right starting point. Pay $20-100/month, focus on the workflows, do not think about infrastructure.
When you should self-host
- You have data residency, compliance, or security requirements.
- Your workflows touch internal systems behind your VPN.
- Execution volume is high enough that cloud per-execution pricing becomes punishing.
- You want to write custom n8n nodes that do not exist publicly.
- You have or can hire DevOps capacity to keep it running.
A minimal self-hosted setup
A serious self-hosted n8n setup needs five things: the n8n container, a PostgreSQL database for state, Redis for queue mode (recommended once you cross ~100 workflows or have long-running executions), a reverse proxy with TLS, and credential encryption.
A single $20/month VPS handles all five, runs hundreds of workflows, and is up in an afternoon if you know your way around Docker. The only ongoing operational work is OS patching and backups.
The traps
Self-hosted n8n in production has a few footguns we have hit and you can avoid:
- Run in queue mode from day one if you expect any long executions. Default mode runs everything in the main process and a single hung workflow can block the whole instance.
- Encrypt credentials at rest. n8n supports this; people forget to set the encryption key in env and discover later that credentials are stored in plain JSON.
- Set up backups for the PostgreSQL database, not just the n8n volume. Workflow JSON, credentials, and execution history all live in the database.
- Pin a version. n8n updates often and minor versions sometimes break custom node compatibility. Pin in your Docker compose, upgrade deliberately.
Hybrid is fine
Plenty of teams run n8n cloud for non-sensitive workflows and self-hosted for the ones touching internal systems. The two can coexist. The "right" answer is not always all-in on one model.
How we deploy n8n for clients
When we set up self-hosted n8n under our n8n automation service, the standard deploy includes: queue mode with Redis, encrypted credentials, automated backups, monitoring with alerts on workflow failures, a workflow-as-code pipeline (workflows version-controlled in git, deployed via CI), and runbooks your team can act on. The infrastructure cost is typically $30-80/month, the engineering setup is a few days.
If you have outgrown Zapier, are spending more than $300/month on Make, or need to bring automation behind your firewall — that is the call to make.