n8n Pricing 2026: Is n8n Free, and What Does Cloud Cost?

Short version. Self-hosted n8n is free permanently, with no cap on executions, workflows or users. n8n Cloud is not free: a 14-day trial, then €20/month billed annually for 2,500 executions, rising to €50 and €667 for the tiers above it.

Both halves of that sentence are true at once, which is why the pricing question is so persistently confusing. Here is the specific contradiction almost everyone hits first, and it is worth clearing up before any numbers.

You read that n8n is free and open source. You go to n8n.io, click sign up, and it offers a 14-day trial. If it's free, why is there a trial?

Because those are two different products:

n8n.io sells n8n Cloud, which is n8n running the software for you on their infrastructure. That is a paid SaaS product with a trial, exactly like any other. There is no free Cloud tier.
The free thing is a separate download you install on your own server. It is not on the signup page because there is nothing to sign up for. You run docker run or install the npm package, and it is yours.

So "n8n is free" and "n8n costs €20/month" are both correct, and which one applies to you is decided entirely by who runs the server. Nothing else about the software changes.

(And "open source" is doing damage in that sentence too. n8n is source-available, not open source. The code is public but the licence restricts commercial use. That is a separate question and it does not affect what anything costs.)

What's on this page

The free version, precisely
What registering unlocks, for free
The licence catch, in short
What n8n Cloud costs
The limits that actually decide your plan
What happens if you blow past your quota
Sizing a plan before you pay for it
What self-hosting actually costs
The $0 hosting routes, and how they fail
When paying is the right call
Who writes n8n pricing guides, and why it matters
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.

This is also the answer to the question that comes up whenever a larger company evaluates n8n: we're 600 people but only a handful build workflows, and it's all internal, so what do we pay? If those few builders are happy sharing one admin login and you don't need SSO or per-team separation, the answer is nothing. Headcount is not a billing input anywhere in n8n's model. What eventually forces a paid tier is governance, not size.

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.

The licence catch, in short

The version that costs nothing comes with a licence that limits what you may do with it commercially, and it catches out exactly the people most likely to be asking this question: freelancers and agencies planning to run automations for paying clients.

n8n is not open source. It's source-available under the Sustainable Use License, which permits use "only for your own internal business purposes or for non-commercial or personal use" and forbids, among other things, "hosting n8n and charging people money to access it."

In practice that means:

Automating your own or your company's operations is free, at any scale and any revenue.
Charging clients for automations is free, as long as the workflows run on the client's own instance.
Running your clients' workflows on your instance requires an Enterprise licence. The arrangement is what matters, not what you charge for it.

That last line is the one that costs agencies money, and the fix is cheap if you apply it before you've onboarded twelve clients onto one server. The full treatment (the three licence conditions, the credential test that decides embedding cases, the OEM and Embed routes for building a product on n8n, and how to get an answer out of n8n in writing) is on n8n commercial use: which licence you actually need.

What n8n Cloud costs

Cloud is n8n hosting n8n for you. There's no free tier, just a 14-day trial first (no card needed on Starter and Pro; the Business trial does ask for one).

Plan	Billed annually	Billed monthly	Executions/mo	Concurrent	Shared projects	Execution history
Starter	€20/mo	€24/mo	2,500	5	1	1 day
Pro	€50/mo	€60/mo	10,000	20	3	5 days
Business	€667/mo	—	40,000	—	6	30 days
Enterprise	Contact sales	—	Custom	200+	Unlimited	365+ days

Paying monthly costs 20% more than paying annually: €24 against €20, €60 against €50. Framed the other way, committing to a year saves you about 17%. Both numbers describe the same gap; people quote them interchangeably and then argue about it.

Every plan includes unlimited workflows and unlimited users. n8n dropped the active-workflow caps that used to restrict the lower tiers, so seat count and workflow count are no longer billing levers at all. Execution volume is.

Business and Enterprise add the governance features: SSO via SAML or LDAP, Git version control, and the option to run the paid tier on your own infrastructure rather than n8n's.

The billing model is genuinely in n8n's favour, and it's the part worth understanding. 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 exactly what one run of a 3-node workflow costs.

Compare that to per-task pricing, where every step bills separately. The same automation can differ by an order of magnitude 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 changes what you end up building, not just what you pay. The Zapier comparison works that math through properly.

The limits that actually decide your plan

Everyone compares execution counts, because that's the number in the big text. For a lot of people it's the wrong number to optimise. Three others bite harder and earlier.

Execution history is 1 day on Starter. This is the one that surprises people in production. A workflow that fails at 11pm on Friday has no retained execution record by Monday morning: no input data, no error detail, nothing to debug from. Pro gives you 5 days, Business 30. If you're running anything you'd have to explain to a client, 1 day of history is a tighter constraint than 2,500 executions, and it's the most common reason to be on Pro rather than Starter regardless of volume.

Concurrency is separate from volume. 2,500 executions with a concurrency limit of 5 is comfortable for scheduled jobs and awkward for a webhook that receives bursts. A spike queues rather than fails, but latency climbs, and a caller with a short timeout will give up while your workflow is still waiting in line. Pro's 20 is a different kind of headroom, not just more of the same.

Shared projects: 1 on Starter, 3 on Pro. Fine solo. Immediately limiting the moment two teams want their own space with their own credentials, which is exactly when someone starts asking about permissions.

What happens if you blow past your quota

Better than most people assume, and worth knowing because the fear of a hard stop pushes people onto plans they don't need.

Per n8n: if you exceed your quota, "your workflows will continue running without interruption, but overage charges may apply." So it doesn't cut you off mid-month, and a traffic spike won't silently break your automations. It also means an accidental infinite loop is a billing event, not just a warning. Business overage runs €4,000 per 300,000 additional executions, so the per-execution rate is roughly in line with the plan itself rather than punitive.

Two practical consequences. Set an alert on your execution count rather than relying on being cut off, because you won't be. And treat a runaway loop in a scheduled workflow as a cost risk on Cloud in a way it simply isn't on self-hosted, where the same bug costs you nothing but CPU.

Sizing a plan before you pay for it

Most people estimate by counting workflows. That's the wrong unit and it undercounts badly. Count trigger frequency, then multiply by sub-workflow depth.

A schedule that fires every 15 minutes is ~2,880 runs a month on its own. That single workflow exceeds the Starter allowance before you add anything else. Every 5 minutes is ~8,640, which is most of Pro.
Sub-workflows each count as their own execution. A parent calling three sub-workflows is four executions per logical run, not one. Well-factored, modular builds burn quota noticeably faster than their node count suggests, and this is the single most common reason a bill comes in higher than expected.
Retries count. A flaky API that fails twice before succeeding is three executions.
Webhooks are the hard one to predict, because the volume isn't yours to control. Size these from the caller's actual traffic, not from your test runs.

Add those up, then look at which row of the table you land in, and check the history column against how much debugging you expect to do, not just the execution column.

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. Two shapes cost far more than people budget for: an AI agent, which makes several model calls per run, and a RAG chatbot, which pastes retrieved documents into every single message.

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

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 Pro rather than Starter when you need the execution history (see above). This is the upgrade people make for the wrong reason and then find they needed anyway.

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 are binary requirements that appear the day someone asks about compliance, not nice-to-haves you grow into gradually.

Pay for an Enterprise or Embed licence when your business model is the thing the licence restricts: you're hosting clients' workflows, or embedding n8n in a product. Which licence you actually need walks through the cases.

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.

Who writes n8n pricing guides, and why it matters

Worth saying plainly, because it changes how you should read every article on this topic including this one.

A large share of the "n8n pricing" results are published by managed-hosting providers, companies whose product is running n8n for you at €5–15/month. Their pricing analysis and their sales pitch are the same document. That doesn't make their numbers wrong, and several are accurate, but it does mean the framing is reliably tilted: Cloud gets presented as expensive, self-hosting as straightforward, and the maintenance burden that would push you back toward Cloud gets a short paragraph near the end.

The mirror image also exists. Competing automation platforms publish "n8n pricing" breakdowns whose conclusion is that you should use their product, and at least one currently ranking is still labelled 2025.

Our own disclosure: the Hetzner, DigitalOcean and n8n Cloud links on this page are affiliate links, marked as such. We earn if you sign up through them, on either side of the decision. That's why the recommendation above is conditional rather than a single answer: it depends on your execution volume and on how much you want to own a server.

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).

If n8n is free, why does the signup show a 14-day trial? Because signing up on n8n.io signs you up for n8n Cloud, which is the paid hosted product. The free version isn't something you sign up for. It's a download you run on your own server. Same software, different party paying for the machine.

Is there a free n8n Cloud plan? No. Cloud starts at €20/month billed annually (€24 monthly) after the trial. Free means self-hosting.

Can I use n8n for free just to learn? Yes, and self-hosting isn't the only route. Running it locally on your own laptop costs nothing and needs no server. Scheduled workflows only run while the machine is on, which is fine while you're learning. Start with the beginner tutorial.

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.

How many executions do I actually need? Count trigger frequency, not workflows, then multiply by sub-workflow depth. A schedule firing every 15 minutes is ~2,880 runs a month by itself, which already exceeds Starter. See sizing a plan.

What happens if I go over my execution limit? Nothing breaks. n8n says workflows "will continue running without interruption, but overage charges may apply." You get billed, not cut off, so set your own alert.

Why is my n8n bill higher than the number of workflows suggests? Almost always sub-workflows. Each one is a separate billed execution, so a parent plus three children is four per run. Retries count too.

Does a bigger plan make my workflows run faster? Not per-run, but higher tiers raise the concurrency limit: 5 on Starter, 20 on Pro. If your workflows are queuing behind each other during bursts, that's the number that helps, not the execution allowance.

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. Which licence you actually need covers the client-work and product cases in detail.

Is n8n open source? No, and it matters here because "free" and "open source" are not the same claim. The code is public, but it ships under a licence that restricts commercial use, which is why the free edition can be free forever without being open source.

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. Check what an export actually carries before you move or share one: n8n workflow templates covers what stays in the JSON.

Does the free version leave out any integrations? No. Every node and every integration is available on Community, on self-hosted, and on all Cloud plans. n8n's pricing page is explicit that pricing is by execution count "regardless of complexity". What money buys is organisational rather than functional: SSO, projects with role-based access, log streaming, external secrets, and Git-backed environments.

Primary sources for the figures above: n8n's pricing page and edition comparison, checked 3 September 2026. Prices are the published euro figures; n8n bills US customers the equivalent dollar amount. 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.

Doing client work? Read which licence you actually need before you consolidate clients onto one instance. That's the arrangement that changes the answer.

Weighing it against a paid tool? n8n vs Zapier pricing works through the cost math. The two bill on different units, which matters more than the headline prices. Brand new to n8n? The beginner tutorial covers the data model first.

Why the signup offers a trial when everyone says n8n is free