n8n Self-Hosted vs Cloud: Where the Math Flips
n8n self hosted vs cloud, priced out: what Cloud execution limits cost, the work self-hosting adds, and the point where a VPS wins.
n8n self hosted vs cloud comes down to one question: do you want to pay per workflow run, or pay for a server and run as many as it can handle? n8n is a tool that links your apps and acts when something happens, like a new order. Its Cloud plans bill per run. The self-hosted Community edition is free, so the bill is the server plus your own time.
What n8n Cloud actually charges for#
An execution is one complete run of one workflow. n8n's own pricing page is clear that the number of steps inside it does not matter. A ten-node flow that runs once costs the same as a two-node flow that runs once. What adds up is frequency.
As listed on n8n's pricing page in October 2026, billed annually:
| n8n Cloud plan | Approximate price | Executions per month | Concurrent runs | Log retention |
|---|---|---|---|---|
| Starter | about 20 euros a month | 2,500 | 5 | 7 days |
| Pro | about 50 euros a month | 10,000 | 20 and up | 30 days |
| Enterprise | quoted by sales | negotiated | 200 and up | 365 days and up |
Monthly billing runs about 17 percent higher. Starter also caps a single run at five minutes, which matters for long scrapes or AI chains.
Now do the arithmetic on one ordinary workflow. A check that runs every five minutes is 288 executions a day, or roughly 8,600 a month. That one schedule eats Starter's allowance three times over and most of Pro's. Add a webhook that fires on every new order and you are shopping for the next tier. Cloud pricing is fair for flows that run a few times a day. It gets expensive fast for anything that polls.
What self-hosting n8n really costs#
The software is free. n8n publishes the Community edition under its Sustainable Use License, which allows self-hosting for your own internal business use. Paid self-hosted tiers exist for features such as single sign-on and embedding n8n in your own product. If you do not need those, there is no execution counter at all.
What you pay instead is a server. n8n runs on Node.js, which pushes most of a workflow through one core at a time, so clock speed matters more than core count. That is why our n8n hosting guide recommends the Performance line: AMD EPYC 4000 series hosts at 4.5 GHz, with every node capped at 60 percent average CPU load. These figures render live from our plan data:
| Fits | Plan | vCPU | RAM | NVMe storage | Monthly |
|---|---|---|---|---|---|
| A handful of personal flows | P10 | 1 | 2 GB | 25 GB | $8.00/mo |
| Most production instances | P20 | 2 | 4 GB | 50 GB | $16.00/mo |
| Heavy schedules, big payloads, AI steps | P40 | 2 | 8 GB | 100 GB | $32.00/mo |
Every one of those plans runs the five-minute schedule above, all day, without a meter. The limit is the CPU and memory on the card, not a count. If you are unsure how much memory your mix of flows needs, our post on how much RAM a VPS needs walks through it.
The work you take on when you self-host#
The server price is the smaller half of the cost. The bigger half is that you become the operator. Here is the list we would hand a friend before they switch.
- A domain and TLS. Webhooks need a public HTTPS address. Put a reverse proxy (a small web server that receives traffic and passes it to n8n) in front of port 5678. Our guide to a reverse proxy with free TLS covers it.
- Updates. n8n ships often. Running the official Docker image makes an upgrade a pull and a restart. Start with Docker and Compose.
- Backups. Your workflows, execution history, and every stored credential live in n8n's database. Lose it and you rebuild from memory. See backing up a Dockerised app.
- Locking the editor down. The n8n editor holds API keys for every app you connect. It should not be open to the whole internet. Our guide to keeping a dashboard private shows how.
- The database choice. n8n starts on SQLite, a single-file database that is fine for getting going. n8n recommends PostgreSQL for high volume or its queue mode, and it runs happily on the same server.
Be honest with yourself about this list. Our VPS plans are unmanaged: we keep the hardware, network, and uptime SLA honest, and you own the n8n install. The difference is spelled out in managed vs unmanaged support.
When n8n Cloud is the right call#
Cloud wins when your flows run a few hundred times a month and nobody on the team wants a server. It also wins when you need n8n's own support contract. Paying a vendor to own upgrades is a reasonable trade at low volume. We made the same argument in reverse in VPS vs cloud hosting: metered billing is fine until your usage stops being small.
Self-hosting wins when any flow polls on a schedule, when execution counts climb past a few thousand, or when credentials should stay on a disk you control. In our Seattle facility, that disk sits on RAID NVMe with one free off-node backup on every VPS plan.
FAQ#
Is self-hosted n8n really free?
The Community edition is free to self-host for your own internal use under n8n's Sustainable Use License. You pay for the server it runs on. Paid self-hosted tiers only matter if you need features like single sign-on.
How many executions can a self-hosted n8n handle?
There is no counter, so the ceiling is your hardware. Long or heavy flows use more CPU and memory than quick ones. When one server gets busy, n8n's queue mode with PostgreSQL spreads the work across worker processes.
What are the minimum requirements for self-hosted n8n?
Light personal use runs on a small server, which is why we list the P10 for it. For production flows we recommend the P20, and the P40 for heavy schedules or AI steps. Fast single cores matter more than many slow ones.
Can I move from n8n Cloud to self-hosted later?
Yes. Workflows export as JSON and import into a self-hosted instance. Credentials do not travel with them, so plan to reconnect each app once on the new server.