n8n on NAS: Run Your Own Self-Hosted Automation Hub

By Leo, woCyber — writing about local-first storage, private cloud and DIY home security.

Here is the rule before you spend anything: n8n starts at about 1 vCPU and 2 GB of RAM, and the software is free when you self-host it. Almost every NAS in a small office already clears that bar. What usually fails is not the processor — it is the workflow database sitting on a disk nobody backs up.

That distinction matters because people buy the wrong thing. They compare CPUs, then lose six months of automations to a single failed volume. n8n on NAS is a storage problem wearing a compute costume.

So the question worth answering is not “can a NAS run it”. It is “which part of a NAS does n8n actually lean on, and what does that cost you over three years”.

Key Takeaways

• Self-hosted n8n needs roughly 1 vCPU and 2 GB of RAM to start, with 2 vCPU and 4 GB for comfortable load, per a 2026 self-hosting guide (Automate All); the container setup itself is documented by n8n’s Docker install guide.
• The software costs nothing under the fair-code Sustainable Use License; internal business use is permitted, reselling n8n as a service is not (Automate All, 2026; the license terms are n8n’s own).
• In a 2026 cost comparison, hosted n8n Cloud’s entry plan was listed at twenty-four US dollars per month with 2,500 executions included, against roughly 30 minutes a month of maintenance when you self-host (Automation Atlas; n8n’s pricing page lists the current tiers).
• Light use, up to about 20 workflows, fits 1 vCPU, 2 GB of RAM and 20 GB of SSD storage (Automation Atlas, 2026).
• n8n ships a new minor version most weeks (Automate All), so whatever you deploy needs a one-command update path — Docker images give you that.

The Short Answer: Can a NAS Actually Run n8n?

Yes, if the NAS can run containers. n8n ships an official Docker image and a documented Docker Compose setup, and a NAS with a container manager is a container host with disks attached.

n8n is a fair-code workflow automation engine: you build a chain of trigger and action nodes in a browser editor, and the server runs that chain on a schedule or on an incoming webhook. Self-hosted, it keeps your workflows, your credentials and your execution history on your own machine.

Two conditions decide whether it works in practice. The NAS must expose persistent storage to the container, so workflows survive a reboot. And someone must own updating the image — the release cadence above is not a suggestion.

So What Actually Limits It — CPU, RAM or Disk?

RAM first, disk second, CPU a distant third.

Most n8n workloads are waiting, not calculating. A workflow that watches a webhook, calls an API and writes a row spends microseconds on the CPU and milliseconds waiting on the network. That is why 2 GB of RAM covers light use: the engine, a small database and a handful of concurrent runs.

Where people get caught is sharing. A NAS that is also indexing a photo library, transcoding video or running an AI service is not the 2 GB machine the documentation describes. My working rule: if the NAS does anything else that is memory-hungry, double the RAM figure you budgeted. That doubling is a rule of thumb, not a published requirement.

Disk is the quieter constraint. The workflow database, credentials and execution logs all grow, and they grow faster than people expect when a workflow runs every minute. Twenty gigabytes is the documented starting point for light use; give the container a share of a real array rather than a USB stick.

Which Route Costs Less Over Three Years?

At that listed twenty-four US dollars a month, three years of the entry cloud plan is eight hundred sixty-four US dollars of recurring spend; a self-hosted stack costs the hardware plus about 30 minutes of maintenance a month. The catch is real, though: that maintenance is you, and a self-hosted instance whose database was never backed up is more fragile than a managed plan, not less.

What Did We Look At, and What Could We Not Check?

• n8n’s own Docker installation documentation — the image, the Compose file and the persistent volume.
• n8n’s pricing page — which capabilities sit behind the paid tiers rather than in the self-hosted build.
• A 2026 self-hosting cost breakdown — server sizing tiers in vCPU, RAM and SSD terms.

We did not run n8n on our own hardware for this article, and we quote no benchmark figures of our own. We also cannot tell you how much RAM a specific NAS will have left over after its own services load, because that number varies per model and per firmware version. Ask the vendor for the free-memory figure at idle; that is the number that decides your container budget.

How Do You Set It Up Without Losing Your Workflows?

1. Confirm the NAS runs containers. Look for Docker, a container manager or a container app package. If there is no container support, stop here and pick a different host.
2. Create a persistent folder on the array for the n8n data directory, and point the container at it. If that folder lives only inside the container, a routine image update erases your workflows.
3. Use PostgreSQL rather than the default SQLite file once more than one person uses the instance or it runs around the clock. SQLite does not handle concurrent writes well.
4. Set the encryption key to something you record outside the NAS. n8n encrypts stored credentials with it, so losing it locks you out of your own integrations.
5. Back up the database, not just the folders. A weekly database dump to a second location on the array, plus one copy off the NAS, is the minimum that survives a failed disk.
6. Do not expose the editor to the internet. Webhooks need to be reachable; the admin interface does not. Put it behind a reverse proxy with authentication, or keep it on your LAN.

Where Does This Break in Real Use?

• The unbacked-up data directory. A tidy one-command update becomes a wipe if the folder was inside the container. Recovery means rebuilding workflows by hand.
• Webhooks published without authentication. An open webhook can be triggered by anyone who learns the URL, and each trigger costs you a run. Our article on keeping AI workloads on private hardware covers the same exposure pattern.
• Storage as an afterthought. Execution history accumulates quietly, then the volume fills and every run fails at once.

Our Take: The Database Is the Product

If you already own a NAS that runs containers, put n8n on it. The marginal cost is a few gigabytes of RAM and you avoid a recurring fee entirely.

The counterintuitive part: the spec that matters is not on the CPU line of any datasheet. It is whether the data directory sits on a backed-up array, and whether the database gets dumped. I have seen more automations lost to an unprotected volume than to a slow processor, and the processor is the thing everyone compares.

Where I would push back on self-hosting: if nobody in the business will own the weekly update and the database dump, pay for a managed plan. An unpatched n8n behind an open webhook is a worse deal than the entry managed plan, and the failure is silent.

For hardware, we ship the storage side rather than the workflow side. Our NAS line — X4 four-bay, S2, A3 — covers the array a container host needs, and the X4 four-bay chassis is the unit most small offices start with. If your deployment needs container or virtual-machine support, tell us the requirement and we will confirm the configuration in writing before you order, because that is a specification decision, not something to assume from a product photo.

What We Don’t Know

We cannot give you a memory ceiling for a specific NAS model, because the free RAM at idle depends on firmware and on which services are enabled. Our doubling rule — 2 GB becomes 4 GB if the NAS does other memory-hungry work — is an assumption, not a measurement. If you need a firmer number, ask the vendor for idle memory usage with your intended services enabled. Treat any self-hosting estimate without that figure as provisional.

Frequently Asked Questions

Do I need a subscription to run n8n on a NAS?

No. The self-hosted build is free under the fair-code licence for internal business use, and n8n’s own pricing page shows the paid tiers as optional. Subscription only becomes part of the picture if you switch to the hosted cloud product or want team features reserved for enterprise plans.

How much RAM does n8n on NAS really need?

Plan on 2 GB for light use, up to about 20 workflows, and 4 GB when the instance runs around the clock or handles concurrent webhooks. If the NAS also runs media indexing, transcoding or an AI service, budget double the figure you first calculated.

Can I run n8n on a NAS without Docker?

Technically you can install it on Node.js directly, but that path means managing runtime versions by hand on a device with a closed update cycle. Docker is the practical route because updates become one command and the runtime stays isolated from the NAS firmware.

What happens when the internet goes down?

Self-hosted automations that stay inside your network keep running, which is the main reason to host locally. Workflows that call an external API pause until the connection returns, and n8n records the failed execution rather than losing it.

One condition before you ask for a quote: container support gets confirmed in writing against the unit you buy, not assumed from a product photo. Then tell us your workflow count, whether the instance must run around the clock, and how much storage you want for execution history — we answer within 24 hours on business days with a configuration priced as a one-time purchase, not a plan.