This comparison is for people who’ve already decided n8n is their automation platform — usually technical founders, developer-adjacent ops people, or small teams who hit the pricing or flexibility walls of Zapier and Make (if you’re still deciding between platforms, start with the three-way comparison). The question now is deceptively simple: run it yourself on a cheap VPS, or pay for n8n Cloud?
I run both. Self-hosted for my own infrastructure, Cloud for two clients who shouldn’t be trusted with SSH keys (their words, roughly). Here’s the honest accounting.
A disclosure worth making: this site earns affiliate commissions, and self-hosting earns us nothing — n8n’s free self-hosted tier has no referral economics. I’m recommending it anyway where it fits, because it genuinely is the right answer for a chunk of readers, and you’ll trust the rest of this site more if I say so plainly.
What you’re actually choosing between
n8n Cloud is the managed SaaS: sign up, build workflows, n8n handles servers, updates, backups, SSL, and uptime. Pricing as of mid-2026 is roughly $24/month (Starter, 2,500 executions) and $60/month (Pro, 10,000 executions), with annual discounts — check the current pricing page, these tiers shift.
Self-hosted n8n is the same application running on hardware you control. The canonical small-business setup is a Docker container on a $6–12/month VPS (Hetzner, DigitalOcean, etc.), with Caddy or Traefik for SSL and a Postgres container beside it. The software license costs $0 under n8n’s fair-code license for internal business use. There’s no execution cap — the limit is your server.
A few features differ: Cloud includes some conveniences (managed OAuth for certain integrations, one-click version upgrades), while a couple of enterprise features (SSO, advanced RBAC, environments) are paid regardless of where you host.
The naive cost comparison
| n8n Cloud Starter | n8n Cloud Pro | Self-hosted (VPS) | |
|---|---|---|---|
| Subscription | ~$24/mo | ~$60/mo | $0 |
| Server | — | — | ~$6–12/mo |
| Executions | 2,500/mo | 10,000/mo | Unlimited* |
| Active workflows | Capped per plan | Higher cap | Unlimited |
| Sticker total/year | ~$288 | ~$720 | ~$72–144 |
*“Unlimited” means “until your VPS runs out of CPU/RAM” — in practice a 2-vCPU/4GB box comfortably handles tens of thousands of executions a month for typical workflows.
Self-hosting looks 5–10× cheaper. The naive table is also wrong, because it prices your time at zero.
The true cost: your time
Here’s what self-hosting actually demands, from someone who does it:
Setup (one-time): 2–6 hours. Docker Compose file, domain + DNS, reverse proxy with SSL, Postgres, environment variables, webhook URL configuration, first backup script. If you’ve deployed Docker apps before, it’s an afternoon. If you’re learning Docker via this project, budget a weekend and some swearing.
Updates: ~20–40 minutes/month. n8n ships releases fast — multiple per month. You don’t need every release, but security patches and breaking changes (they happen; read the release notes) mean a monthly docker compose pull && docker compose up -d ritual, with a backup first, and occasionally a migration that needs attention.
Backups: 1 hour setup, then vigilance. Your workflows, credentials, and execution history live in that Postgres database. No backups = one bad disk away from rebuilding everything. A nightly pg_dump to object storage is cheap insurance; you must actually set it up and occasionally test a restore.
Security: ongoing, and it’s all yours. Your n8n instance stores OAuth tokens and API keys for every service it touches — your CRM, your email, your payment provider. Exposed to the internet (it must be, for webhooks), it’s a target. You own: keeping the host patched, not exposing the Postgres port, strong auth on the n8n UI (and ideally 2FA or an access proxy), and watching for CVEs. n8n the company patches the software fast; applying patches is on you.
Incidents: rare but real. Disk fills with execution logs, container restarts after a host reboot misconfigured, an update changes webhook behavior. Budget a couple of hours a quarter for “why is my automation not running” archaeology — at exactly the moment the automation failing is costing you.
Monitoring: the part everyone skips. Cloud tells you when something’s wrong; self-hosted tells you nothing unless you build the telling. At minimum you want an uptime check on the n8n URL (free tiers of any uptime monitor work), a disk-space alert on the host, and an n8n error workflow that pings you on Slack or email when any production workflow fails. That’s another hour of setup — and the difference between finding out from a dashboard versus finding out from an angry customer three days later.
Putting numbers on it
Call it 1–2 hours/month of maintenance after setup, for a competent operator. The comparison then becomes:
| Your effective hourly value | Self-hosted true cost/mo (VPS + 1.5h) | Cloud Pro cost/mo | Cheaper option |
|---|---|---|---|
| $30/hr | ~$55 | ~$60 | Roughly a wash |
| $75/hr | ~$120 | ~$60 | Cloud |
| $150/hr | ~$235 | ~$60 | Cloud, clearly |
| You’d do the sysadmin anyway / enjoy it | ~$10 | ~$60 | Self-hosted |
That last row is not a joke — it’s the most common real case. If you already run a VPS for other services, the marginal cost of adding an n8n container is genuinely near zero, and the “maintenance hours” overlap with work you’re doing anyway. Model your own numbers in the automation cost calculator.
When self-hosting clearly wins
- You already operate servers. Marginal effort ≈ zero, savings are pure.
- High execution volume. At 50,000+ executions/month, Cloud pricing climbs while a $20 VPS yawns. Heavy AI agent workflows — like a busy version of our AI customer support bot — rack up executions fast.
- Data residency / compliance. Customer data never leaves infrastructure you control. For EU clients with strict processors lists, this ends the conversation.
- You need community nodes, custom nodes, or local LLMs. Self-hosted n8n can talk to an Ollama instance on the same box; that’s a uniquely self-hosted superpower.
- Tinkering tolerance is high and budget is tight. Bootstrapped founders: $720/year buys a lot of groceries.
When Cloud clearly wins
- You’re the only technical person and you’re the bottleneck. Every maintenance hour comes out of product or sales. Pay the $24–60 and go build revenue.
- The automations are business-critical and you have no on-call habit. n8n Cloud’s uptime and managed backups beat a hobbyist VPS the one night it matters.
- Team use without an ops culture. Letting three non-technical teammates build workflows on a server nobody maintains is how you get a security incident with extra steps.
- You want managed OAuth and one-click upgrades rather than reading release notes.
Try it yourself
n8n
n8n Cloud starts around $24/month — all the power of n8n with zero server babysitting. Try it before you commit to running your own.
Start with n8nThe hybrid path most people actually take
The pattern I recommend to most readers: start on n8n Cloud, migrate later if the math says so. Workflows export as JSON and import into a self-hosted instance nearly unchanged (credentials must be re-entered — they’re encrypted, sensibly, and don’t travel). You learn the platform with zero ops burden, and by the time your bill or volume justifies self-hosting, you’ll know n8n well enough that the Docker setup is the easy part.
The reverse migration also works, which makes this a low-regret decision either way. Compare that to Zapier or Make, where your workflows are hostages of the platform — n8n’s portability is itself a feature neither hosting choice takes away.
Verdict by persona
- Developer / technical founder with existing infrastructure: self-host. It’s an afternoon of setup for ~$600/year of savings and total control. This is the correct answer for you and no affiliate incentive changes that.
- Technical-ish founder, no servers yet, automations becoming critical: n8n Cloud Starter. Revisit self-hosting when you outgrow 10,000 executions or hire someone who likes ops.
- Small team, mixed skills: Cloud. The managed upgrade path and not having a single point of sysadmin failure is worth the subscription.
- Agency deploying n8n for clients: self-host per client (or one multi-tenant box if you know what you’re doing) — but only if maintenance is a service you bill for. Free server-tending is how agency margins die.
Whichever side you land on, you’ve already made the bigger correct decision: a platform where your workflows are portable, your costs are inspectable, and the ceiling is your skill rather than your plan tier.