Self-Hosted vs Cloud n8n: What You Trade for Control

Self-hosting n8n looks cheaper until you count the hours. Cloud n8n looks expensive until you count the hours you get back. That is the whole debate in one sentence, and most of the online arguments about it miss the point because they treat the decision as ideological instead of operational.

n8n is one of the few automation platforms that actually gives you a real choice: run it yourself on your own infrastructure, or pay them to run it for you. Both work. Both are production-grade. The question is which set of costs you would rather pay, and that answer depends on things most vendor comparisons refuse to name.

What you actually get from n8n Cloud#

n8n Cloud is a managed instance. You sign up, pick a plan based on workflow executions, and start building. Updates happen automatically. Backups happen automatically. When something breaks in the underlying infrastructure, someone else pages themselves at 2 AM, not you.

The pricing shakes out to roughly $20 to $50 per month for the starter tiers, scaling with execution volume and active workflows. For a small business running a few dozen workflows a month, that is coffee money. For a company running hundreds of thousands of executions with heavy data throughput, the bill starts to sting.

What Cloud gives you that people undervalue: SSO on higher tiers, isolated execution environments, log retention you did not have to configure, and a support channel that exists. What it takes away: filesystem access, the ability to install arbitrary npm packages inside function nodes, and any illusion that your data never touches someone else's servers.

If you are automating client-facing workflows, invoicing, or anything with light PII, Cloud is almost always the right starting point. You are not saving money by self-hosting until you are big enough that the savings exceed the labor.

What self-hosted n8n really costs#

The sticker price on self-hosted n8n is zero. The Community Edition is free, open source, and will run on a $6 VPS if you want it to. That number is misleading.

Here is the actual cost sheet for a serious self-hosted deployment:

• Infrastructure: $20 to $200 per month depending on load, redundancy, and whether you run a separate Postgres instance and Redis for queue mode.
• Setup time: 4 to 20 hours to get a production-ready environment with HTTPS, backups, monitoring, and a reverse proxy that does not fall over.
• Ongoing maintenance: 2 to 8 hours per month for updates, log rotation, credential rotation, dependency patches, and the occasional broken workflow after a version bump.
• Incident response: unpredictable. When your instance goes down at 6 AM on a Tuesday, you are the on-call engineer.

Multiply those hours by whatever your time is worth, or the time of the person you are paying to do it. For most small businesses running under 50,000 executions per month, self-hosting costs more in true labor than Cloud costs in cash. The math flips somewhere between 100,000 and 500,000 executions, depending on how complex your workflows are and how much your ops person costs.

What self-hosting actually buys you: complete data sovereignty, no per-execution pricing anxiety, the ability to run community nodes and custom code without restriction, and direct database access when you need to debug or migrate. Those are real benefits. They are not free.

The three questions that actually decide it#

Forget feature checklists. The decision comes down to three questions, and if you answer them honestly you will know which path to take.

Does your data have to stay on infrastructure you control? If you handle healthcare records, financial data under specific regulatory regimes, or client contracts with data residency clauses, self-hosting is often not optional. Cloud is fine for a lot of regulated work, but "often fine" is not the same as "contractually acceptable." Read your obligations before you read pricing pages.

Do you have someone who can own the instance? Not somebody who can install Docker once. Somebody who will patch it, monitor it, rotate credentials, respond when it goes down, and stay current on n8n's release notes. If that person does not exist inside your business and you are not willing to pay for one, do not self-host. A neglected self-hosted n8n is a liability, not an asset.

Is your execution volume high enough that Cloud pricing hurts? Pull your projected workflow count and multiply by realistic execution frequency. If the answer puts you comfortably inside a $50 to $200 per month Cloud plan, the labor of self-hosting is not worth the savings. If you are staring at a $1,500 per month Cloud bill, self-hosting starts to make sense as a line item.

If you answered no, no, and no, use Cloud. If you answered yes to at least two, self-hosting deserves serious consideration.

The middle path most people ignore#

There is a third option that gets almost no attention: managed self-hosting. You run n8n on your own infrastructure, but you pay someone else to handle the ops layer. This can mean a fractional DevOps contractor, a platform like Elestio or Railway that hosts n8n for you with more control than Cloud offers, or a consulting arrangement where a firm manages the instance on your behalf.

This path costs more than raw self-hosting and less than the labor of doing it yourself. It preserves most of the data control benefits. It is the right answer for businesses that need self-hosting for compliance reasons but do not have internal engineering to spare.

The tradeoff: you are now dependent on that provider or contractor. If they disappear, you inherit the instance cold. Make sure whoever manages it documents everything and gives you real access, not a black box.

What people get wrong about the switch#

Two common mistakes worth naming.

First, businesses migrate from Cloud to self-hosted to "save money" without accounting for labor, then quietly migrate back a year later after two outages and a credential leak. If you cannot articulate the labor budget before you start, you are not saving money. You are deferring costs into a category you refuse to measure.

Second, businesses self-host from day one because it feels more serious, then spend the first three months fighting infrastructure instead of building automations. Cloud is not a beginner tier you graduate from. It is a legitimate production choice that lets you focus on the actual work. Plenty of mature businesses stay on Cloud forever because the math never flips.

The right sequence for most small businesses: start on Cloud, build enough workflows to know what you actually need, then reassess after six months of real usage data. You will make a better hosting decision with numbers than with hypotheses.

The honest recommendation#

Cloud for almost everyone under 100,000 monthly executions with no strict data residency requirements. Self-hosted for regulated industries, high-volume operations with dedicated ops capacity, or teams that need custom code and community nodes as a core part of their workflows. Managed self-hosting for the middle case where compliance matters but internal engineering does not exist.

Whichever you pick, the workflows are the value. The hosting is plumbing. Do not spend six weeks optimizing the plumbing before you have workflows worth plumbing for.

If you want help sizing your n8n deployment against your actual volume and compliance needs, see how we approach it.