Contents
- Fast answer
- Quick facts
- What this covers
- Why it matters
- Where to go next
- Outcomes
- Step by step
- The methodology: infrastructure, then the hours nobody counts
- The hours threshold, and why it is the only number that matters
- Sizing the server: concurrency, not monthly volume
- The limits: five costs that never appear on any invoice
- When to self-host regardless of the arithmetic
- At a glance
- Mistakes to avoid
- FAQ
- Related guides
Fast answer
Self-hosting n8n costs the price of a small server plus the hours you spend on it, and the second half is what decides whether it beats n8n Cloud. The n8n Community edition is free on GitHub with no execution metering at all, so a six to forty-eight dollar VPS, backups at a percentage of that, a domain and some monitoring is the entire invoice. n8n Cloud publishes Starter at twenty euro for 2,500 executions, Pro at fifty euro for 10,000 and Business at 667 euro for 40,000, with seventeen percent off on annual billing. The calculator on this page converts those euro figures at a dated reference rate, adds your setup and ongoing hours at your own rate, and then answers the only question that matters: how many hours a month can you afford to spend before Cloud is cheaper. For most people the threshold is small enough to be sobering.
Facts last verified 2026-08
Quick facts
- n8n Community edition
- Free on GitHub, no execution metering
- Cloud Starter
- EUR 20/mo, 2,500 executions
- Cloud Pro
- EUR 50/mo, 10,000 executions
- Cloud Business
- EUR 667/mo, 40,000 executions, self-hosting permitted
- Annual billing
- 17% cheaper, per n8n's pricing page
- Server range used
- $6 to $48/mo DigitalOcean Basic droplets
- Prices checked
- 2026-08-20
On this page
What this page covers
A tool visitor should leave with a decision, not just a number: build now, prepare first, choose another workflow, or follow a course path.
- 01Execution volume
- 02Server and backups
- 03Your time priced
- 04Cloud tiers
- 05What money misses
- 06Sources
Why does this matter now?
Self-hosting is presented as the frugal choice and it usually is, right up until the person who set it up is on holiday during a failed upgrade. The infrastructure genuinely is cheap: n8n meters executions rather than steps, and the Community edition does not meter at all, so the marginal cost of another ten thousand runs is zero. What is not free is the upgrade cadence, the backup you have never restored, the credentials sitting on a box you are now responsible for, and the bus factor of one. None of those appear on an invoice, which is exactly why they get left out of the comparison and exactly why the comparison keeps coming out wrong. Pricing your own hours turns a slogan back into a decision.
Internal path
Where to go next from this page
These links are part of the A8gent learning and conversion path. Use them to move from concept, to diagnosis, to workflow build, to course.
What you should be able to do after this
- Total self-hosted monthly cost
- The equivalent n8n Cloud tier
- Cost per execution both ways
- The ops-hours threshold where Cloud wins
How do you do it, step by step?
1. Count executions, not steps
n8n bills one execution per workflow run whatever the node count, on Cloud and Community alike. This is the structural reason n8n gets relatively cheaper than step-metered tools as workflows grow, and it means your volume estimate only needs one number. Pull it from your existing instance or from your trigger frequency.
2. Size the server for concurrency, not volume
Executions per month tells you very little about the box you need; simultaneous executions and memory-hungry nodes tell you everything. Start on a small droplet, watch memory under real load, and move up only when you see pressure. If you need queue mode with separate workers, add them in the calculator, because each one is another server at the same price.
3. Add backups, database and the small recurring costs
Backups are commonly priced as a percentage of the server cost, so the calculator expresses them that way and lets you override it. If you move Postgres off the box to a managed instance, that is a separate line and often a larger one than the server. Domain, TLS and uptime monitoring are small but real, and leaving them out is how a twelve dollar estimate becomes a twenty dollar bill.
4. Price setup once and spread it
Standing up n8n behind a reverse proxy with TLS, persistent storage, a real database and working webhooks is a few hours of focused work if you have done it before and a weekend if you have not. Put in an honest number, spread it over the months you expect to run this instance, and note that a short amortisation window is the honest one for a first attempt.
5. Price the ongoing hours, then read the threshold
Version upgrades, credential rotation, the occasional restart, and reading logs when a workflow silently stops. The calculator turns your Cloud plan price into the number of hours a month you can spend before Cloud wins. Compare that threshold to the hours you actually budgeted. If your budgeted hours exceed the threshold, self-hosting is a preference rather than a saving, which is a perfectly reasonable thing to choose knowingly.
6. Check which features you are actually giving up
n8n lists self-hosted deployment as a Business and Enterprise entitlement, and Business and Enterprise features need a paid licence key even on your own hardware. Free self-hosting and fully featured self-hosting are different products. Establish which one you are comparing before you draw a conclusion about price.
The methodology: infrastructure, then the hours nobody counts
The self-hosted column sums six lines. Server cost is your droplet price multiplied by one plus the number of queue-mode workers, because each worker is another instance at the same price. Backups are a percentage of that server cost, defaulting to twenty and editable, because that is how backup pricing is commonly expressed. A managed database is a separate figure you supply, defaulting to zero for an on-box Postgres. Domain, TLS and monitoring are a small flat figure. Then ongoing ops hours at your rate, and one-off setup hours at your rate divided by the months you spread them over.
The Cloud column is a lookup. The calculator finds the cheapest published tier whose execution allowance covers your monthly volume, converts the euro list price at a dated reference rate, and applies a 0.83 factor if you toggle annual billing, which is the seventeen percent saving n8n states on its own pricing page. Past 40,000 executions there is no published tier, so the page says quote rather than inventing an extrapolation.
The comparison then does something most cost pages avoid: it solves for your time rather than for the money. Subtract infrastructure and amortised setup from the Cloud plan price, divide by your hourly rate, and you get the number of hours a month you can spend on the box before Cloud is the cheaper option. That single figure is the honest form of the question, and it is uncomfortable in a productive way. If the threshold is 1.4 hours and you budgeted three, self-hosting is a preference, not a saving.
Cost per execution is shown for both columns because it makes the volume effect obvious. The self-hosted figure falls continuously as executions rise, since nothing in that column moves with volume. The Cloud figure falls within a tier and then steps up at each tier boundary. Seeing the two curves in the same units explains, in one glance, why self-hosting arguments get stronger as volume grows and weaker as hourly rates grow.
The hours threshold, and why it is the only number that matters
Work a case. Ten thousand executions a month puts you on Cloud Pro at EUR 50, about $58.03. Self-hosted on a $24 droplet with twenty percent backups is $28.80, plus three dollars of domain and monitoring: $31.80 of infrastructure. Six hours of setup at eighty-five dollars spread over twelve months adds $42.50 a month. That is $74.30 before you touch it once, which already exceeds the Cloud plan, and the page will tell you plainly that infrastructure and setup alone have lost the argument.
Change the amortisation to thirty-six months and setup drops to $14.17, so the pre-maintenance total is $45.97 and the threshold becomes about twelve dollars of your time, or roughly eight and a half minutes a month at eighty-five dollars an hour. Eight and a half minutes. That is the honest picture at Pro-tier volume for someone with a professional hourly rate, and it is why so many self-hosting decisions look better in a spreadsheet than in a year.
Now move the volume. At 40,000 executions Cloud Business is EUR 667, about $774, and the same self-hosted setup on a larger droplet is still under a hundred dollars including several hours of maintenance. The threshold becomes hours per month rather than minutes, and self-hosting is decisively correct on cost. The lesson is not that self-hosting is good or bad but that it has a volume at which it starts winning, and that volume is much higher than the enthusiasm suggests.
The third lever is your rate. At eighty-five dollars an hour the threshold is minutes; at twenty dollars an hour it is hours; if the person maintaining it is genuinely idle otherwise, the threshold is effectively infinite. Be honest about which of those describes you, because the calculator will faithfully compute whichever answer you feed it.
Sizing the server: concurrency, not monthly volume
The most common sizing mistake is reading the monthly execution count and picking a droplet from it. Monthly volume tells you almost nothing. What determines the box is how many executions run at the same moment and how much memory the heaviest node holds. A workflow that fires once an hour and processes a large file needs more machine than one that fires two hundred times an hour and moves three fields.
The practical approach is to start small, watch memory under real load for a fortnight, and move up when you see pressure rather than when a number crosses an arbitrary line. n8n's Cloud tiers publish concurrency rather than hardware, which is a useful hint about what actually constrains the system: Starter lists five concurrent executions and Pro twenty. If your workload needs more simultaneity than that, that is the signal to think about queue mode.
Queue mode splits the main instance from separate worker processes, and in this calculator each worker is another instance at the same droplet price, which is the right first approximation. It is worth understanding that queue mode buys throughput and resilience, not lower cost. Adding two workers triples your server line and does not reduce a single execution charge, because there were never any execution charges on Community in the first place.
If you move Postgres to a managed instance, budget for it as its own line and expect it to be larger than the application server. That is usually a good trade, since a database you are not personally backing up is one fewer thing that can quietly cost you a year of work, but it changes the arithmetic materially and the calculator has a field for it precisely so it does not get forgotten.
The limits: five costs that never appear on any invoice
Version upgrades. Community releases move quickly, and the cost of upgrading is small and recurring while the cost of not upgrading is large and lumpy. A skipped quarter becomes a migration, and a migration is a weekend rather than an hour. None of this shows up in the monthly figure, and all of it shows up in the year.
Restore, not backup. Paying twenty percent for snapshots is not the same as having a recovery process. A backup that has never been restored is a belief. The first real failure on an untested backup costs more than several years of the Cloud plan, and that is a comparison worth making before rather than after.
Credentials. Every API key your workflows use now lives on a machine you are responsible for securing, patching and access-controlling. For many organisations that is fine and for some it is a compliance obligation with real work attached. Either way it is a cost, and it is one that the hosted option quietly absorbs.
Bus factor. If exactly one person understands the docker-compose file, the reverse proxy config and where the environment variables live, self-hosting is cheap right up until that person is unavailable during an incident. Writing the runbook is the mitigation, and writing the runbook is also hours nobody put in the estimate.
Licensed features. n8n lists self-hosted deployment as a Business and Enterprise entitlement, and enterprise capabilities require a paid licence key even on your own hardware. Free self-hosting and fully featured self-hosting are different products at different prices, and comparing the free one against a paid Cloud tier that includes those features is not a like-for-like comparison.
When to self-host regardless of the arithmetic
Data residency and network reach are the two reasons that legitimately override cost. If workflows must touch systems on a private network, or if data may not leave infrastructure you control, the calculator becomes documentation of what that requirement costs rather than an argument about whether to meet it. That is a perfectly good use of the page: knowing the price of a constraint is useful even when the constraint is not negotiable.
Volume is the third reason, and unlike the other two it is visible in the numbers. Because Community does not meter executions at all, the self-hosted cost per execution keeps falling while the Cloud cost steps up at each tier. Past the published tiers the comparison stops being close, which is why high-volume operators self-host and low-volume ones generally should not.
If you are going to self-host, budget the setup properly and write the runbook while you still remember what you did. The single highest-return hour in the whole project is the one spent documenting the restore procedure, because it converts the largest unpriced risk on this page into a known quantity. The long-form guide to the true cost of self-hosted n8n covers what that runbook should contain.
And if the threshold this page produced was smaller than the hours you honestly expect to spend, take the Cloud plan and spend the hours on workflows instead. That is not a defeat. Deciding to pay somebody else for undifferentiated maintenance is the same reasoning that led you to automate the work in the first place.
At a glance
Published n8n Cloud tiers and DigitalOcean droplet prices used by this calculator, read on 2026-08-20. Euro converted at 1 EUR = 1.1605 USD, the ECB reference rate for 2026-08-19.
| Item | Price per month | Capacity | Notes |
|---|---|---|---|
| n8n Cloud Starter | EUR 20 (about $23.21) | 2,500 executions | 5 concurrent executions, 1 shared project |
| n8n Cloud Pro | EUR 50 (about $58.03) | 10,000 executions | 20 concurrent executions, 3 shared projects |
| n8n Cloud Business | EUR 667 (about $774) | 40,000 executions | Self-hosted deployment permitted, 6 shared projects |
| n8n Enterprise | Quoted | Custom | 200+ concurrent executions, hosted or self-hosted |
| n8n Community edition | $0 | Unmetered | Free on GitHub; enterprise features need a paid licence key |
| DigitalOcean 1 vCPU / 1 GB | $6 | n/a | 25 GB SSD |
| DigitalOcean 1 vCPU / 2 GB | $12 | n/a | 50 GB SSD |
| DigitalOcean 2 vCPU / 4 GB | $24 | n/a | 80 GB SSD |
| DigitalOcean 4 vCPU / 8 GB | $48 | n/a | 160 GB SSD |
| Annual billing discount | 17% off | n/a | Stated on the n8n pricing page for all Cloud plans |
What mistakes should you avoid?
- Comparing the server price against the Cloud plan price and calling that the total cost of self-hosting.
- Setting ongoing maintenance hours to zero, which is the assumption that decides the outcome and never survives a year.
- Running without tested backups, so the first real failure costs more than several years of the Cloud plan.
- Undersizing the box for concurrency because the monthly execution count looked small.
- Skipping upgrades until a skipped upgrade becomes a migration.
- Assuming self-hosting unlocks every feature, when some capabilities require a paid licence key regardless of where it runs.
FAQ
- How much does it cost to self-host n8n?
- The software is free. The bill is a VPS, which on DigitalOcean Basic droplets runs six, twelve, twenty-four or forty-eight dollars a month depending on size, plus backups, optionally a managed database, and a few dollars for domain and monitoring. Call it roughly ten to sixty dollars a month of invoices for a single instance. The real total is that figure plus your hours, which is what the calculator on this page exists to add.
- Is self-hosted n8n cheaper than n8n Cloud?
- On infrastructure alone, usually yes, and increasingly so as execution volume rises because the Community edition does not meter executions. Once you price your own time it is much closer, and the calculator states the threshold explicitly: the number of hours a month you can spend before Cloud is the cheaper option. At a professional hourly rate that threshold is often only a couple of hours.
- What are the n8n Cloud prices?
- As published in August 2026, Cloud Starter is twenty euro a month for 2,500 executions with five concurrent executions, Cloud Pro is fifty euro for 10,000 executions with twenty concurrent, and Cloud Business is 667 euro for 40,000 executions with self-hosted deployment permitted. All tiers include unlimited active workflows and unlimited steps per execution, and n8n states annual billing saves seventeen percent. Enterprise is quoted rather than published.
- What size server do I need for n8n?
- Concurrency and node memory drive this far more than monthly volume. A small instance handles light workloads comfortably; workflows that process large files, run headless browsers or hold big datasets in memory need considerably more. Start small, watch memory under real load, and move to queue mode with separate workers when a single instance stops keeping up rather than when the execution count crosses an arbitrary line.
- What does self-hosting give me that Cloud does not?
- Control over where data lives, the ability to reach systems on a private network, no execution metering, and the freedom to run any version you like. Those are real advantages and for some organisations they are the whole reason. They are also independent of price, which is why a self-hosting decision made purely on cost tends to be the one that gets reversed.
- What is the biggest hidden cost?
- The backup you have never restored, followed closely by the bus factor. A backup that has not been tested is a belief rather than a backup, and an instance only one person knows how to rebuild is cheap right up to the week that person is unavailable. Neither appears in any monthly figure on this page, and both are larger than the server bill when they land.
Sources & further reading
Was this page helpful?
