Is n8n Free? What's Free, What Costs Money, and the License Catch (2026)
n8n's self-hosted Community edition is free with no execution limit — but the licence restricts commercial use. What you get free, what Cloud costs, and what self-hosting really costs.
By Ali Ilyas · · Updated July 30, 2026
Short answer: yes, if you self-host it. The Community edition of n8n is free to download and run, with no cap on workflows, executions, or users, and no time limit. n8n Cloud is not free — there's a 14-day trial, then plans start at €20/month.
But "free" is doing two jobs in that sentence, and only one of them is about money. The version that costs nothing also comes with a licence that limits what you're allowed to do commercially — and that limit catches out exactly the people most likely to be asking this question: freelancers and agencies planning to run automations for paying clients.
This page covers both: what the free version actually includes, and where the free licence stops.
What's on this page
- The free version, precisely
- What registering unlocks, for free
- n8n is not open source
- Can you use free n8n for commercial work?
- The freelancer and agency question
- Building a product on top of n8n
- What n8n Cloud costs
- What self-hosting actually costs
- The $0 hosting routes, and how they fail
- When paying is the right call
- Frequently asked questions
The free version, precisely
Self-hosted n8n comes in editions. The default one — what you get when you run the Docker image or install the npm package — is the Community edition, and it's free permanently, not as a trial.
What you get:
- No execution quota. n8n doesn't meter executions on a self-hosted instance. Your ceiling is your server, not a billing plan.
- Unlimited workflows, active workflows, and users in the same sense: nothing counts them.
- Every node and every integration, including the AI and LangChain nodes.
- Almost the complete feature set. n8n's own wording.
What's held back for paid self-hosted tiers, per n8n's edition comparison:
| Feature | Community | Paid (Business / Enterprise) |
|---|---|---|
| Workflow and credential sharing | — | ✅ |
| Projects | — | ✅ |
| SSO (SAML, LDAP) | — | ✅ |
| Environments | — | ✅ |
| Git version control | — | ✅ |
| External secrets | — | ✅ |
| Log streaming | — | ✅ |
| Multi-main mode (HA) | — | ✅ |
| External binary-data storage | — | ✅ |
Read that list again, because the shape of it matters more than its length. Nothing there is a capability limit — you can build anything on Community that you can build on Enterprise. Everything on the list is about multiple people sharing one instance safely: permissions, secrets, environments, audit trails, high availability. A solo builder loses nothing. A 40-person company with compliance requirements loses the things it can't operate without.
If you haven't got an instance running yet, the self-hosting guide covers the setup end to end, and the Docker guide covers the container specifics.
What registering unlocks, for free
There's a middle tier people miss: registered Community. You enter an email address, n8n sends a free licence key, and three more features switch on at no cost:
- Folders — organise workflows into folders instead of one flat list.
- Debug in editor — copy and pin execution data from a past run back into the editor while you're fixing it.
- Custom execution data — save, search, and annotate metadata against executions.
It costs nothing but an email address. Debug in editor alone is worth it — pulling a real production payload back into the editor is the difference between reproducing a bug and guessing at it.
n8n is not open source
This confuses people, and it's the root of most of the licence confusion further down, so it's worth being exact.
n8n's source code is public on GitHub. n8n is still not open source, and its docs say so plainly. The project calls itself fair-code: n8n's own definition is software that is "source-available, but restrict[s] commercial use of their source code."
The licence is the Sustainable Use License. It grants "the free right to use, modify, create derivative works, and redistribute" — with conditions that a genuine open-source licence, by definition, would not have.
Practically: you can read the code, patch it, fork it, and run your patched version. You cannot do whatever you like with it commercially. Which brings us to the part that actually matters.
Can you use free n8n for commercial work?
Yes — and this is where most people's worry is misplaced. Making money with n8n is fine. Making money from n8n is not.
The licence permits use "only for your own internal business purposes or for non-commercial or personal use." Internal business purposes is a broad allowance, not a hobbyist carve-out. n8n's own examples of permitted use include:
- "Using n8n to sync the data you control as a company"
- Providing consulting services related to n8n
- Creating integrations between n8n and other products
So a company automating its own operations on free Community n8n is fine, at any scale and any revenue. A consultant charging for n8n expertise is fine.
What the licence forbids is anything where "the value derives entirely or substantially from n8n functionality":
- "Hosting n8n and charging people money to access it"
- "White-labeling n8n and offering it to your customers for money"
And separately, a condition that's easy to trip over while doing something otherwise legitimate: "You may not alter, remove, or obscure any licensing, copyright, or other notices of the licensor in the software." Stripping the n8n branding out of the editor because it looks nicer in a client demo is a licence violation on its own, independent of what you charge.
The freelancer and agency question
This is the real question hiding behind "is n8n free," and it comes up constantly: I build automations for clients. Can I run them on my own instance?
n8n's help centre answers it directly, and the answer turns on whose instance the workflows run on — not on what you charge, how you invoice, or whether you call it hosting:
| What you're doing | Licence you need |
|---|---|
| Helping clients set up and run their own n8n instances | "No commercial license would be required on your part" |
| Hosting and managing your clients' workflows and credentials inside your own instance | "An Enterprise license would be required" |
| Embedding n8n inside your product, white-labelled | An Embed licence — a separate commercial agreement |
That first row is the important one, and it's more generous than the community usually assumes. Building, deploying, maintaining, monitoring, and charging a monthly retainer for automations that run on infrastructure your client owns needs no licence at all. You can charge whatever you like for that work.
The second row is where the trouble is. Consolidating twelve clients onto one instance you own — the natural, efficient, obvious thing to do — is the arrangement that requires an Enterprise licence. Convenience for you is precisely what converts it into reselling access.
The pattern that stays compliant is straightforward, if less tidy: one instance per client, on infrastructure in the client's name, with billing in the client's name. You get access to administer it. A small VPS runs €5–8/month, which the client pays directly. You charge for setup and a maintenance retainer, which is the part with the margin anyway.
Two side benefits, both real. The client's credentials stay in the client's instance, so you're not holding a dozen companies' API keys in one place you'd have to disclose after a breach. And it's cleanly transferable — when the engagement ends you hand over admin access, rather than negotiating an extraction from a shared instance.
None of this is legal advice. The licence text is short and worth reading in full, and n8n answers licensing questions at license@n8n.io. If your arrangement sits near a line, ask them and keep the reply — an email from the licensor is worth more than a forum consensus.
Building a product on top of n8n
If you're using n8n as the engine behind your own SaaS rather than as a tool you operate for clients, the rules are different and the docs are specific.
n8n as an invisible backend — your users never see any n8n UI, workflows run under your service's credentials, n8n is one component of a product whose value doesn't come substantially from n8n — is, per n8n's docs, "available on all paid plans under the standard license, with no separate agreement needed." Note paid plans: this route needs a subscription, not the free Community edition.
Surfacing the n8n editor inside your product so your users build their own workflows is an OEM deployment, and it "requires a separate commercial agreement with n8n." Under an OEM agreement, "n8n branding is required" — your users will see it as n8n embedded in your product, not as your own feature.
Fully white-labelled — your branding, no visible n8n — is the Embed licence, also a separate agreement. Pricing for both isn't published; it's a sales conversation.
The test to apply to your own idea: if your users log into n8n, or supply their own third-party credentials to it, or your product would have nothing left without it, you're in agreement territory. If n8n is plumbing your users never encounter, you're on the standard licence with a paid plan.
What n8n Cloud costs
Cloud is n8n hosting n8n for you. There's no free tier — a 14-day trial (no card required on the Starter and Pro trials), then:
| Plan | Price (billed annually) | Executions/month | Concurrent executions |
|---|---|---|---|
| Starter | €20/month | 2,500 | 5 |
| Pro | €50/month | 10,000 | 20 |
| Business | €667/month | 40,000 | — |
| Enterprise | Contact sales | Custom | 200+ |
Paying monthly instead of annually costs roughly 17% more. Starter and Pro include unlimited users and unlimited workflows.
The billing model is the part worth understanding, and it's genuinely in n8n's favour. n8n charges per execution: "An execution is a single run of your entire workflow. It doesn't matter how many steps are in the workflow." One run of a 40-node workflow costs the same as one run of a 3-node workflow.
Compare that to per-task pricing, where every step in the workflow bills separately. The same automation can differ by an order of magnitude in cost between the two models, and the gap widens the more thorough your error handling and data cleanup are. On per-task billing, adding a validation step costs money forever; on n8n's model it costs nothing. That's a real difference in what you end up building, not just what you pay.
Two things worth knowing before you size a plan:
- Concurrency is separate from volume. 2,500 executions with a concurrency limit of 5 is fine for scheduled jobs and awkward for a webhook that receives bursts. A traffic spike queues rather than fails, but latency climbs.
- Sub-workflows count. If you've factored a workflow into a parent and three sub-workflows, that's four executions per logical run, not one. Heavily modular designs consume quota faster than their node count suggests.
What self-hosting actually costs
Free software, not free operation. The honest total:
A server: €5–8/month. Hetzner's ~€6/month 2-core / 4GB tier is the community's default recommendation and comfortably runs a single instance. Hostinger and DigitalOcean land in the same range. 1GB of RAM works for light use, but n8n plus a database plus a Node process is tight, and it's the usual cause of instances that die during large runs.
A domain: ~$10–15/year. Not strictly required, but you want a stable HTTPS URL. Webhook providers register your callback address once, so an address that changes means re-registering it everywhere — see the triggering guide for what that failure looks like from the inside.
API costs, which are usually the biggest line. n8n itself is free; OpenAI, Anthropic, transcription, and enrichment APIs are not. An AI workflow's monthly bill is normally dominated by tokens, not hosting — and the OpenAI integration guide covers the usage limits to set before your first production run.
Your time, which is the cost people leave out. Updates, backups, TLS renewal, disk space, and the occasional 2am debug. Call it a couple of hours a month once it's stable, and more than that while it isn't.
So self-hosting is cheaper than Cloud, not free — and the gap only becomes decisive at volume. At 2,000 executions a month, €6 of VPS versus €20 of Starter is a rounding error against your time. At 200,000 executions a month, the same €6 VPS versus the Cloud plan that covers it is the whole argument.
The $0 hosting routes, and how they fail
You can get to literally zero, and the two common routes both have a specific failure mode worth knowing before you rely on them.
An always-free cloud VM. Oracle Cloud's always-free tier is the community favourite — genuinely free, genuinely permanent, enough resources to matter. The catch is reported often enough to treat as expected behaviour rather than bad luck: free-tier instances are reclaimable, and people lose them without notice. One user's summary was blunt — they stopped the free VPS, and every workflow went with it. The mitigation people report is upgrading to pay-as-you-go with a card on file: the same free allowance applies, you're still billed nothing if you stay inside it, but instances are no longer preemptible. Set a $1 billing alert and you'll know immediately if something falls outside the allowance.
Your own machine. A spare mini PC, an old laptop, a home server. Free, fully under your control, and fine for personal automations — as long as you accept that scheduled workflows only run while the machine is awake and online.
Two traps to avoid regardless of route:
- Free app-hosting tiers that sleep. A host that idles your container after inactivity will silently break every Schedule Trigger you own. People work around it with a cron job that pings the instance every few minutes, which works, and is also a good sign you've picked the wrong host.
- No backups. This is what turns a reclaimed instance from an annoyance into a loss. Export workflows and credentials on a schedule to storage that isn't on the same box. Every "I lost everything" story shares this root cause, not the hosting choice.
When paying is the right call
Not a sales pitch — the specific conditions where free stops being the cheaper option:
Pay for Cloud when the automation matters more than the infrastructure is interesting to you. If a broken workflow costs a client relationship, someone has to own uptime, TLS, and updates. €20/month is less than an hour of your time.
Pay for Business or Enterprise when you hit something on the feature list, and you'll know which one: SSO because IT requires it, Projects and sharing because multiple teams need separate spaces, Git version control because you need review before production, multi-main mode because downtime isn't acceptable. These aren't nice-to-haves you grow into gradually — they're binary requirements that appear the day someone asks about compliance.
Pay for an Enterprise or Embed licence when your business model is the thing the licence restricts. Not because enforcement is likely, but because a licensing problem discovered during acquisition due diligence, or when a client's legal team reviews your stack, is discovered at the worst possible moment.
Stay on free Community when you're a solo builder, a small team happy to share one admin login, or a company automating its own operations. That covers most people asking, and it stays true indefinitely.
Frequently asked questions
Is n8n free forever, or is it a trial? The self-hosted Community edition is free permanently, with no execution limit. Only n8n Cloud has a trial (14 days).
Is there a free n8n Cloud plan? No. Cloud starts at €20/month after the trial. Free means self-hosting.
How many workflows can I run on the free version? Unlimited, and unlimited active ones. n8n doesn't count workflows or executions on a self-hosted instance — the limit is your server's resources.
Can I use n8n for free commercially? For your own or your company's internal purposes, yes, at any scale. What's restricted is selling access to n8n itself — hosting it for paying users, or white-labelling it.
Can I charge clients for n8n automations? Yes, provided the workflows run on the client's own instance. Building, maintaining, and charging a retainer for that needs no licence. Running their workflows on your instance is what requires an Enterprise licence.
Is n8n open source? No. It's source-available under the Sustainable Use License, which n8n describes as fair-code. The code is public; commercial use is restricted.
Do I need to pay to use the AI and LangChain nodes? Not to n8n — they're in the free Community edition. You'll pay whichever model provider you point them at, and note that a ChatGPT subscription doesn't cover API calls: see connecting OpenAI.
What does registering the Community edition give me? A free licence key that unlocks folders, debug in editor, and custom execution data. Email address only, no payment.
Is self-hosting actually cheaper than Cloud? Cheaper in cash, not free. Budget €5–8/month for a VPS plus your own maintenance time. At low volume the difference is small enough that Cloud often wins; at high volume self-hosting wins clearly.
Can I move from Cloud to self-hosted later, or back? Yes. Workflows export and import as JSON. Credentials don't transfer — you re-enter those, and any OAuth redirect URIs need re-registering against the new address.
Primary sources for everything above: n8n's Sustainable Use License, the edition comparison, and the pricing page. Licensing questions go to license@n8n.io.
Ready to run it free? Start with the self-hosting guide, then the Docker setup. Once it's up, the integration guides for Google Sheets, Telegram, and Slack get you to a working automation, and our templates page has starter workflows to import.