Self-hosting n8n is free. The software costs nothing and runs unlimited executions. So the only real question is n8n hosting: where do you run it, which option is best, and what does it actually cost per month?

I went through eight providers and priced each one against its live pricing page in July 2026. Not the headline number on the homepage, the number you pay once n8n has enough RAM to not fall over and a disk that survives a restart.

Best n8n Hosting for Most People

For most people the best n8n hosting is InstaPods at $7/mo flat: 2 vCPU, 2 GB RAM, 25 GB SSD, deployed in about 30 seconds with HTTPS, backups, and WEBHOOK_URL already set up, and nothing to administer. It is a predictable flat price with real resources and zero server admin, which is what most people actually want from hosted n8n.

Being honest about price: PikaPods is cheaper at $3.70/mo, about $3.30/mo less. If absolute lowest cost on light workflows is your only axis, PikaPods wins on the number and I say so below. But it bills off a prepaid balance that ticks down hourly, scopes resources tightly to the app, and gives you no shell. For a flat bill, more headroom, and a real Linux server you can log into, InstaPods is the better default.

Quick picks:

The full table, with what each provider is genuinely best for, is below. The rest of the post is the honest per-provider math behind these picks, and the raw-VPS answer everyone gives on Reddit got a lot worse in June.

n8n Hosting Compared: Price, Specs, and Who Each Is Best For

Sorted by who each provider is best for, not by raw price. The cheapest number is not the same as the best fit, so read the “Best for” column before the price.

Provider Best for Config that actually runs n8n Real monthly cost You administer the server?
InstaPods Best overall / most people Build: 2 vCPU, 2 GB, 25 GB SSD $7 flat, month to month No
PikaPods Lowest price, light workflows n8n minimum resources From $3.70 No
Hetzner (DIY) Running the server yourself CX23 shared vCPU €5.49 (~$6) + your time Yes
Hostinger Most RAM per dollar (if you prepay) KVM 1: 1 vCPU, 4 GB, 50 GB NVMe $6.49 on a 24-month prepay, $11.99 on renewal Yes
Elestio Managed with a support contract Managed n8n, Level 1 support From $11 (Level 2 support is +$50/service) No
Railway Best deploy UX (usage-billed) Hobby + ~1 GB RAM running continuously ~$10-15 (usage-based) No
DigitalOcean (DIY) DIY with better docs and panel 2 GB, 1 vCPU, 50 GB $12 + your time Yes
Render Existing Render users (not the cheap pick) Standard: 2 GB, 1 CPU (Starter is 512 MB) $25 + persistent disk No
n8n Cloud (for reference) Execution limits, not servers Starter, capped at 2,500 executions €20/mo billed annually No

How I priced these: every number above was checked against the provider’s own live pricing page in July 2026, not a third-party roundup. Where a provider bills by usage rather than a flat fee (Railway, PikaPods), I sized the config to n8n’s own documented recommendation of 2 GB RAM and priced that. Where a promo price requires a multi-year prepay (Hostinger), I list both the promo and the renewal, because the renewal is what you actually pay in year three.

What n8n Actually Needs Before You Compare Prices

Half the “n8n costs $4/mo” advice online is priced against a box that will run out of memory the first time you process a large batch.

n8n’s own docs recommend 2 GB of RAM minimum, with 4 GB if you want to be comfortable. Heavy executions spike to 1-2 GB on their own. A 512 MB plan will boot n8n and then die under any real workflow, which is why Render’s $7 Starter tier is not on the table as a serious option even though the price looks great.

You also need persistent storage. n8n keeps its workflows, credentials, and execution history in a database on disk. Any platform where the filesystem resets on redeploy will lose your workflows. Every provider in the table above handles this, but it is the first thing that breaks on the platforms I left out.

So the honest floor for running n8n is: 2 GB RAM, a disk that survives restarts, and a public HTTPS URL so webhook triggers can reach you. Price anything below that and you are pricing a toy.

The Cheap-VPS Answer Got 144% Worse in June 2026

The standard Reddit answer to “cheapest n8n hosting” is “just get a Hetzner box.” That advice is several years stale.

Hetzner raised cloud server prices on 15 June 2026, and the increases were not small:

Hetzner plan Old price/mo New price/mo Increase
CX23 €3.99 €5.49 +38%
CPX22 (2 vCPU, 4 GB, 80 GB) €7.99 €19.49 +144%
CCX13 (dedicated vCPU) €15.99 €42.99 +169%

US regions took it harder still: CPX11 in Ashburn and Hillsboro went from €5.99 to €17.49, a 192% jump. Those are Hetzner’s own published figures, and they apply to new orders and rescales from 15 June onward.

The cheapest Hetzner tier is still genuinely cheap at €5.49. But the moment you want the 2-4 GB of headroom n8n likes, the “obviously cheapest” option is now €19.49, which is more than double what a managed n8n pod costs. If you built your cost model on a €7.99 CPX22 last year, rebuild it.

And that price is before your time. A DIY VPS means you install Docker, write a compose file, set up a reverse proxy, get a TLS certificate, configure WEBHOOK_URL correctly so triggers fire, set up backups, and then patch the box every month. That is a few hours up front and a recurring chore forever. If your time is worth anything, a €5.49 server is not a $5.49 solution. We broke that trade down in full in managed vs self-managed VPS, including the cases where managed is actually the cheaper line item.

What Does It Cost to Keep n8n Running, Not Just to Start It?

Running n8n costs more than the server. n8n shipped 47 releases in the 30 days to 7 August 2026, so version upgrades are recurring work. On a DIY VPS you also own TLS renewal, OS patching, and backups of your workflow data. On a managed pod those jobs belong to the host and the monthly price is the whole price.

That release count is from n8n’s own GitHub releases feed, and it is the number the price table above cannot show you. Installing n8n is an afternoon. Keeping it online is a standing job. Here is the actual list of that job, and who does each part.

Job that exists after n8n is online DIY VPS (Hetzner, Hostinger, DigitalOcean) Managed pod (InstaPods, $7/mo)
n8n version upgrades You. Stop the service, install the new version, hope it starts, restore by hand if it does not One click. A snapshot is taken first, and if n8n does not come back healthy the pod is rolled back automatically
HTTPS certificate renewal You. Certbot plus a renewal timer you only find out has broken when the browser warns someone Ours, at the server level. You never touch certbot
Backing up workflows, credentials, execution history You write it, and you test the restore Daily by default, 5 copies kept on the $7 Build plan. n8n's data directory is in the backup set
OS and Docker patching You, monthly, forever Ours
Public webhook URL and reverse proxy You configure Nginx or Caddy and set WEBHOOK_URL correctly, then re-check it after any domain change Preconfigured at deploy

If you enjoy that work, a €5.49 Hetzner box is a fine answer and always will be. If you do not, the honest comparison is not “$7 versus €5.49”, it is “$7 versus €5.49 plus an hour or two a month that you will sometimes skip until something breaks.”

What Managed Hosting Does Not Fix

Two of the loudest complaints about running n8n in production are not hosting problems, and no host solves them. Anyone who tells you otherwise is selling.

Credential sprawl. API keys, OAuth tokens, and provider configuration spread across n8n’s credential store, environment variables, and whatever else you run, until nobody is sure which key is where or which ones are still live. That lives inside n8n and inside your own key hygiene. Moving the server to a managed host does not consolidate any of it.

Human approval steps. If an AI agent in your workflow can issue a refund, send a quote, or change supplier bank details, you want it to pause and wait for a person before it fires. That is a workflow-design problem you solve inside n8n with a pause step and an approval channel you build. It is not something a hosting plan can hand you.

So the honest scope of what managed hosting buys you is the five rows in the table: upgrades, certificates, backups, patching, and the proxy. It is a real amount of work and it is worth $7. It is not everything. The same breakdown, from our side, is on the n8n hosting page.

Provider by Provider

InstaPods: $7/mo flat, 2 vCPU and 2 GB (best for most people)

Full disclosure, this is us. n8n on InstaPods is $7/mo flat on the Build plan: 2 vCPU, 2 GB RAM, 25 GB SSD, deployed in about 30 seconds with HTTPS, backups, and WEBHOOK_URL already configured.

Where it wins, and why it is the best pick for most people, is predictability and headroom: a flat monthly number with no prepaid balance, no usage meter, no 24-month commitment, and no renewal cliff. You get 2 GB of RAM (n8n’s actual recommendation, not a 512 MB tease) and 2 vCPU, plus SSH and a shell on a real Linux server if you want to poke at it. Managed PostgreSQL, MySQL, and Redis are available if you outgrow the bundled SQLite.

Where it loses: PikaPods is $3.30/mo cheaper and a bare Hetzner CX23 is about a dollar cheaper. If absolute lowest cost is the only axis you care about, we are not the answer, and the next section says so plainly.

PikaPods: cheapest, from $3.70/mo

PikaPods is the price winner, plainly. n8n starts at $3.70/mo on their app listing, and you get a managed instance with no server to look after. If your only axis is the lowest possible number and your workflows are light, this is the pick.

The catch is the billing model and the ceiling. PikaPods runs on a prepaid balance that ticks down hourly rather than a flat monthly charge, with a $1 per pod per month minimum. You watch a balance instead of paying one predictable number. Resources are also tightly scoped to the app, so there is less room to grow into and no shell access to the box.

So PikaPods wins the price line outright and I am not going to pretend otherwise. What you trade for that $3.30/mo saving is the flat bill, the headroom, and the shell. The n8n on InstaPods vs PikaPods breakdown covers where that ceiling starts to bite.

Hostinger: $6.49 that becomes $11.99

Hostinger sells a self-hosted n8n VPS with n8n preinstalled. KVM 1 is 1 vCPU, 4 GB RAM, 50 GB NVMe at $6.49/mo, and KVM 2 is 2 vCPU, 8 GB at $8.79/mo. On specs alone that is a lot of RAM for the money.

Read the term. That $6.49 requires a 24-month upfront prepay, and it renews at $11.99/mo. KVM 2 renews at $14.99. So the real long-run price is roughly double the advertised one, and you prepay two years to get it.

The other thing to be clear about: this is a VPS with a one-click installer, not a managed service. You still own the OS, the updates, the firewall, and the backups. The setup is easier than raw Hetzner. The ongoing maintenance is identical. Side by side detail: n8n on InstaPods vs Hostinger.

Railway: $5 plan, then the meter runs

Railway’s Hobby plan is $5/mo and includes $5 of usage credit. That reads as “$5 n8n hosting”. It is not.

Railway bills RAM at $10 per GB-month and CPU at $20 per vCPU-month, with volumes around $0.25 per GB-month. n8n running continuously at roughly 1 GB of RAM burns through the $5 credit on memory alone, before CPU or storage. Realistically you land at $10-15/mo, and it moves with your workload.

Railway is a genuinely nice platform and the deploy experience is excellent. But usage-based billing on an always-on service like n8n means you are paying for 730 hours a month of uptime no matter what, without the discount a flat plan gives you for that commitment.

Render: $25 once you size it properly

Render’s Starter web service is $7/mo, which looks competitive until you check the specs: 512 MB RAM and 0.5 CPU. That is a quarter of what n8n’s docs ask for.

The tier that fits n8n is Standard at $25/mo for 2 GB RAM and 1 CPU, plus a persistent disk billed separately per GB. That makes Render roughly 3.5x the cost of a $7 flat plan for the same 2 GB of memory, and more expensive than n8n Cloud’s Starter tier.

Render is a good platform for the things Render is for. Always-on self-hosted apps at the cheap end are not it.

Elestio: $11/mo and up, with paid support tiers

Elestio offers fully managed n8n from $11/mo, which covers a dedicated VM, backups, SSL, updates, and monitoring. You pick the underlying cloud provider and instance type, so the final price varies with what you choose.

The part people miss is the support tiers. Level 1 support is free and means community guides and self-service. Level 2, where Elestio troubleshoots and fixes things for you, is $50 per service per month. If you are buying managed hosting specifically so someone else fixes it at 2am, your real price is $61/mo, not $11.

Elestio is a serious product for teams that want a dedicated VM and a support contract. It is not the cheap option and does not try to be. Full comparison: n8n on InstaPods vs Elestio.

DigitalOcean: $12 for the DIY route with better docs

A DigitalOcean Basic droplet with 2 GB RAM, 1 vCPU, and 50 GB storage is $12/mo. Bump to 2 vCPU and it is $18.

That is double Hetzner’s entry tier for equivalent specs, and you are still doing all the same DIY work. What you get for the premium is better documentation, a friendlier control panel, and more regions. For a lot of people that is a fair trade, but it is not a cheap n8n option.

So What Should You Actually Pick?

Two more managed hosts worth a look if none of the above fit: n8n vs Sliplane and n8n vs Sevalla.

The gap between the cheapest option and the most expensive one here is $3.70 versus $25 for the same software doing the same work. That is a 6.8x spread driven almost entirely by billing model, not by capability.

One number that puts the whole table in perspective: n8n Cloud’s Starter plan is €20/mo and caps you at 2,500 executions. Every self-hosted option on this list runs unlimited executions. Even the most expensive one, Render at $25, gives you more throughput headroom than the €20 Cloud plan. The reason to self-host n8n is not really the monthly saving, it is that the execution meter goes away entirely.

FAQ

What is the best n8n hosting in 2026? For most people the best n8n hosting is InstaPods at $7/mo flat: 2 vCPU, 2 GB RAM, 25 GB SSD, deployed in about 30 seconds with HTTPS, backups, and WEBHOOK_URL already configured, and nothing to administer. It is a flat, predictable price with real resources and no server admin. If your only priority is the lowest possible number on light workflows, PikaPods is cheaper at $3.70/mo, with the trade-offs below.

What is the cheapest way to host n8n in 2026? PikaPods at $3.70/mo is the cheapest managed option. A Hetzner CX23 at €5.49/mo is the cheapest DIY option, though you administer the server yourself. For a flat, predictable price with 2 GB of RAM and no administration, InstaPods is $7/mo, which is the pick most people are better off with.

Is self-hosted n8n free? The software is free. n8n Community Edition is source-available under the Sustainable Use License with unlimited executions and no license fee. You pay only for the server it runs on, which is $3.70-12/mo depending on the provider you pick. See n8n pricing in 2026 for the full Cloud versus self-hosted math.

Is self-hosting n8n worth the maintenance? It depends on whose maintenance it is. n8n shipped 47 releases in the 30 days to 7 August 2026, so upgrades are a standing job, and on a DIY VPS you also own TLS renewal, OS patching, and backups. On a managed pod at $7/mo those five jobs are the host’s: upgrades are one click with an automatic snapshot and rollback, certificates renew on our side, and workflows back up daily. What no host fixes is credential sprawl and human-approval steps inside your workflows - those are n8n-level problems.

Should I just write a Python or Node script instead of running n8n? Often, yes, and it is worth saying that plainly. If you are a developer with one trigger and one or two APIs to call, a cron job and forty lines of code is less to run than a whole automation service. n8n earns its keep in three specific cases: you have many integrations to authenticate against and do not want to write each client, someone who is not a developer needs to read or change the flow, or you need visual execution history to debug the run that failed at 3am. If none of those are true, write the script and keep one less service alive.

How much RAM does n8n need? n8n’s docs recommend 2 GB minimum and 4 GB to be comfortable. Heavy executions spike to 1-2 GB. Avoid 512 MB plans, which rules out Render’s $7 Starter tier.

Is Hostinger’s $6.49 n8n hosting really that cheap? Only for the first two years, and only if you pay those two years upfront. It renews at $11.99/mo. It is a VPS you administer, not a managed service.

Can I move my n8n instance between providers? Yes. n8n workflows export as JSON and the whole instance is portable. Migrating n8n walks through moving from Cloud or another host without losing credentials or execution history.

What if I want a different automation tool entirely? Price is rarely the reason to switch tools, but if n8n is the wrong fit, the 9 best self-hosted n8n alternatives compares Activepieces, Windmill, Node-RED and others by license and RAM footprint.

Where n8n Lands Against the SaaS Tools

If you got here because n8n itself is the cheap alternative to something else, those comparisons are broken out separately: n8n vs Zapier ($7/mo versus $20/mo with per-task limits), n8n vs Make (unlimited operations versus per-operation billing), and n8n vs Power Automate ($7/mo flat versus $15 per user).

Each of those comparisons assumes the $7/mo self-hosted number from the table above. That is the point of pricing this properly: the “cheap” side of every n8n comparison holds only if your hosting bill stays where you think it is.

Ready to skip the setup? Deploy n8n on InstaPods for $7/mo flat. 2 vCPU, 2 GB RAM, HTTPS and webhooks configured, running in about 30 seconds.

All prices verified against each provider’s live pricing page in July 2026. Providers change prices without notice, and Hetzner’s June 2026 adjustment is a good reminder to re-check before you commit to a multi-year term. Release-cadence figures updated 8 August 2026 from n8n’s GitHub releases feed.