n8n hosting: what each route costs, and what breaks first
Three routes, one decision: who is the sysadmin, and what happens when the workflow count grows.
Every figure below was read from the provider’s own pricing, documentation or legal page on 2026-09-12. Where a provider does not publish something, the table says so instead of guessing.
Short answer first
Zero maintenance: the managed cloud starts at 20 €/month billed annually for 2,500 workflow executions a month, with unlimited users and workflows, and a free trial that needs no credit card.
Software free, you run the server: the documented floor is 2 vCPU / 4 GB RAM. On that spec you pay $24/month at DigitalOcean, $11/month at IONOS on a one-year term (first three months $4), or €5.50/month for the first 24 months at Contabo for 4 vCPU / 8 GB. The n8n licence itself is $0 — the Community edition is the same software, self-hosted.
You already own the hardware: the licence is still $0, but you supply the memory, and 4 GB is the floor the documentation states. Hardware prices vary too much to publish a single number; see the notes under the table.
Not published anywhere we could verify (2026-09-12)
Hetzner’s current cloud price list (its pricing blocks render dynamically), Vultr’s and Linode/Akamai’s spec-level prices (same reason), Oracle’s Always-Free compute specification, the entry plan’s trial length, and a bare-board hardware price. These are marked “not published” rather than filled in with a number we could not attribute.
The comparison table
| Option | Starting price | Minimum spec | Free tier / trial | Hard limits | Data date |
|---|---|---|---|---|---|
| Managed cloud — entry tier | 20 EUR/mo, billed annually | fully managed (no spec to choose) | Free trial, no credit card required (length not stated) | 2,500 executions/mo; 5 concurrent; 2.5k saved executions; 2.5 GB saved-execution storage; 7-day log retention; SQLite | as of 2026-09-12 |
| Managed cloud — mid tier | 50 EUR/mo, billed annually | fully managed | same free trial | 10,000 executions/mo; 20 concurrent; 25k saved executions; 25 GB storage; 30-day retention; SQLite | as of 2026-09-12 |
| Managed cloud — business tier | 667 EUR/mo, billed annually | fully managed (self-hosted option) | same free trial | 40,000 executions/mo; SSO/SAML/LDAP; 6 shared projects; 30-day insights | as of 2026-09-12 |
| Managed cloud — enterprise tier | not published (contact sales) | fully managed or self-hosted | same free trial | Custom executions; 200+ concurrent; unlimited retention; log streaming; SLA | as of 2026-09-12 |
| Self-host — Community edition (software) | $0 (free download) | 2 vCPU / 4 GB RAM minimum per official guidance | n/a — the software itself is free | No execution cap — hardware is the cap; default DB is SQLite; no official support | as of 2026-09-12 |
| Self-host on DigitalOcean | $24.00/mo (2 vCPU / 4 GiB / 80 GiB SSD) | 2 vCPU / 4 GiB | new-account credit promos: not published on the pricing page checked | 4,000 GiB outbound transfer included | as of 2026-09-12 |
| Self-host on IONOS (VPS M+) | $11/mo on 1-year term ($4/mo first 3 months) | 4 vCores / 4 GB RAM / 120 GB NVMe | 30-day money-back guarantee (refund; not a free tier) | Unlimited traffic; promo price requires 1-year term | as of 2026-09-12 |
| Self-host on Contabo (Cloud VPS 4) | 5.50 EUR/mo for first 24 months | 4 vCPU / 8 GB RAM / 100 GB SSD | not published | “Unlimited traffic” — fair-use policy permits throttling | as of 2026-09-12 |
| Own hardware (mini-PC / NAS / SBC) | not published (device-dependent) | 2 vCPU / 4 GB RAM per official guidance | n/a | Home network and port forwarding; no SLA; electricity; ARM images vary | as of 2026-09-12 |
| Oracle Cloud Free Tier | $0 for Always Free services | exact free compute shape: not published on the page served | Always Free services indefinitely + USD 300 credit for 30 days; one account per person | Capacity limits; production workloads expected to move to paid resources | as of 2026-09-12 |
What the table does not tell you: what breaks first
Prices are the easy part. These are the failure points that decide whether the cheap row stays cheap.
Memory is the binding constraint, not CPU. n8n’s own deployment guide treats CPU casually — “for most usage levels, a basic shared CPU plan is enough” — but the same page, for the AI sandbox runner that uses Docker-in-Docker, states the floor explicitly: provision at least 4 GB RAM / 2 vCPU. The sandbox runner is the part that needs headroom, and it is the part most people enable after the first week.
SQLite gets you further than you expect, and then stops. Self-hosted installs use SQLite by default and it is genuinely fine for a single user. It stops being fine in three specific situations: (1) execution history grows inside the same file, so the database bloats with log data; (2) writes lock the whole database file, so parallel work serialises; (3) the moment you switch on queue mode — the only mode the documentation describes as scalable — SQLite is no longer an option at all, because multiple worker processes write concurrently. Postgres and Redis become hard requirements at that point, not optimisations. Even the managed cloud runs SQLite on its entry and mid tiers, with PostgreSQL reserved for the enterprise scaling plans. If your plan is to grow into queue mode, plan the database from day one.
Queue mode has a second requirement nobody mentions in the pricing table: the filesystem binary-data mode is not supported in queue mode. If workflows persist binary data, that storage has to move to S3-compatible object storage first.
Retention is a limit, not a setting, on the cloud. The entry tier keeps 7 days of execution logs and 2.5 GB of saved executions. Self-hosters set that themselves (there is a pruning setting), and the ones who never touch it are the ones who discover their VPS disk filling up six months later.
How to choose, in one minute
- You want it working today and you value your evenings: entry managed tier, accept the execution cap.
- You want the software free and you’ll keep a box updated: 2 vCPU / 4 GB VPS, Postgres from the start, backups on.
- You want to scale workers or run heavy AI/browser nodes: budget 8 GB, add Redis and object storage, expect the database migration.
Sources
- n8n cloud pricing and self-hosting documentation (read 2026-09-12)
- DigitalOcean, IONOS, Contabo and Oracle product/pricing pages (read 2026-09-12)