n8n Cloud Pricing: Worth It Over Self Hosted?

Updated 14 September 2026 · n8n, automation, pricing, self-hosting

The honest answer on n8n cloud pricing is that the paid plan is worth it over self hosted when your automations touch someone else's money, and rarely worth it before that. As of September 2026, secondary pricing guides put the Cloud Starter plan at €24 a month for 2,500 executions, while the self-hosted Community edition carries no licence fee at all. That gap of roughly €288 a year buys you one thing: somebody else is awake when the container dies at 2am.

Most articles on this keyword stop at the price table. The table is the easy part, and every one of those guides has it. What none of them answer — and what no source gathered for this article answers either — is how many hours of your own time the €24 is buying back, and at what point in an automation practice that trade flips. There is no published, usage-based return-on-investment figure for freelancers specifically; the reasoning below is this article's own, built from the prices and limits that are sourced, not from a study that measured it.

The prices, and who published them

Here are the numbers as multiple 2026 buyer guides report them, plus a column that says plainly which of them n8n's own pricing page actually confirms.

Here is where each number in that table comes from, starting with the one source that isn't a secondary guide: n8n's own pricing page. That page, checked 14 September 2026, shows a Starter price of 20€/month billed annually for 2,500 executions with unlimited steps, and describes Pro as built for solo builders and small teams. It does not show monthly-billing prices, Business or Enterprise pricing, or any concurrency or workflow-history numbers in the sections available for this article — those gaps are marked "No" above, not filled in with a guess.

For the numbers n8n's own page doesn't show, this article relies on 2026 buyer guides, and they don't all agree with each other. CloudZero, on 10 September 2026, reports Starter at €24/month, Pro at €60/month and Business at €800/month, all billed monthly, plus an annual discount of roughly 17% without stating exact annual prices per plan. CheckThat.ai's pricing page, dated 30 March 2026 and last updated 13 September 2026, independently lands on the same Starter and Pro numbers — €24/€20 monthly/annual for Starter, €60/€50 for Pro — without citing CloudZero, which is a stronger signal than either guide alone. ConnectSafely.ai, dated 8 September 2026 and updated 10 September 2026, reports the same three monthly figures and a free, licence-less self-hosted tier.

Two of the guides disagree outright. Instapods, dated 24 July 2026 and updated 9 September 2026, reports the same Starter and Pro executions but adds a claim none of the others make: that "every Cloud plan now includes unlimited workflows and unlimited users." That sits awkwardly next to Arahi.ai's update of 10 July 2026, which reports a 5-active-workflow limit and a 320 MiB RAM ceiling on Starter specifically. Those two claims cannot both be current. n8n's own page doesn't settle it in the sections reviewed here, so treat the workflow-count and RAM limits on Starter as unresolved rather than picking a side.

The concurrency and workflow-history columns — 5/20/30/200+ concurrent runs, 1/5/30/365 days of history — come from a single guide, ToolRadar, updated 4 September 2026, and aren't corroborated anywhere else in the sources gathered for this piece. Treat them as the least certain numbers in the table. The Business annual price of roughly €667/month isn't published by any guide directly; it's this article's own arithmetic, applying CloudZero's ~17% annual discount to the €800 monthly Business price. AutomateNexus, updated 3 September 2026, reports €20 and €50 as the Starter and Pro annual prices, matching CheckThat.ai. The self-hosted Community edition's zero licence fee is confirmed on n8n's own pricing page; the claim that it places no cap on executions or workflows follows from how self-hosted software works — there's no vendor meter on your own server — and is also how SyncGTM's review, updated 26 August 2026, describes it when it says "no stated cap."

Every figure above was checked on 14 September 2026, and pricing pages move, so open the official page before you put a card in. No source gathered for this article — this one included — has a current n8n Cloud invoice in hand; what's here is other people's reading of the vendor's page, cross-checked against each other and against the page itself where it says anything at all.

What "unlimited steps" actually changes about the maths

The single most useful thing in n8n's pricing model is that a Cloud execution is one workflow run, however many nodes it contains. The vendor's page says Starter covers 2,500 executions with unlimited steps. That is a different unit from what Zapier charges on, and it is why people migrate.

Work it through with a real shape of workflow. Say you have an inbox triage flow: a Gmail trigger, a classifier call to an LLM, a switch node with four branches, a Slack notification, a row appended to a sheet, and an error branch. That is eight or nine nodes. On a per-task platform you are billed for most of those every time an email arrives. On n8n Cloud, one email is one execution. At 2,500 executions on Starter, you can process roughly 83 emails a day inside the cheapest paid plan, no matter how baroque the workflow gets.

This is the thing that makes the €24 look cheap and it is genuine. It is also the thing that makes people overestimate their headroom, because executions are consumed by triggers you forgot about. A schedule trigger set to every five minutes burns 8,640 executions a month on its own, which blows past Starter and most of Pro before a single useful thing happens. Polling triggers are the usual culprit. If you are coming from a tool where polling was free, check your trigger intervals before you check your plan.

The hidden cost of self hosting, priced in hours

Self hosting costs €0 in licence and somewhere between two and six hours a month of attention, which is the number the price tables never show.

Here is what that time is actually spent on, based on how the software works rather than on invented test logs:

• Server and database. n8n needs somewhere to run and a Postgres database behind it for anything beyond toy use. A small VPS that will not fall over under a queue-mode setup runs a few dollars a month, so infrastructure is not the expense. Standing it up, locking down the port, and putting a reverse proxy with TLS in front of it is an evening for someone who has done it before, and a weekend for someone who hasn't.
• Upgrades. n8n ships new versions frequently. Every upgrade is a container pull, a restart, and a check that workflows still run. Nodes change, and a workflow built against an older node version can behave differently after an upgrade. On Cloud this happens without any of that manual checking.
• Backups that have actually been tested. An untested backup is a feeling. Workflows, credentials and execution history all live in that database, and losing the credentials store means re-authorising every OAuth connection by hand.
• Webhook reachability. Any workflow triggered by an external service needs a public HTTPS endpoint that stays up. This is the piece that breaks most home-lab n8n installs without any warning — the failure mode is silent: the webhook just never arrives.
• Being the on-call engineer. If the instance is down at 2am and a client's order confirmations run through it, whoever owns the box is the person who fixes it. There is no support queue.

Put a number on an hour of freelance time. At $60/hour, three hours a month of maintenance is $180 of that time against a €24 bill. That is the entire argument in one line, and it is why the answer for most working freelancers is Cloud, even though the software itself is free.

The counter-argument is just as real. For anyone who already runs servers, already has monitoring, and already does upgrades on other infrastructure, the marginal cost of adding n8n to that stack is close to zero. Three hours a month becomes twenty minutes because it rides along with work that was happening anyway. Sysadmins who self-host n8n are making the right call for their situation.

When self hosted is clearly the better call

Self hosting wins outright in four situations, and in each one the reason is structural rather than financial.

High-volume, low-value executions. The Community edition is reported to place no cap on executions, and a workflow that fires 50,000 times a month doing something trivial doesn't fit the Cloud execution ladder well. That volume sits above the reported 40,000 on Business, and Business is reported at €800 monthly.

Data that cannot leave a given piece of infrastructure. Health records, legal documents, or anything under a client contract that names where data may be processed makes self hosting the only version of this conversation. No pricing argument beats a compliance requirement.

Learning the tool. Spending a month inside a self-hosted instance teaches more about how n8n actually works than a year of clicking around a managed one, because the queue, the database and the logs are all visible. The wider version of this question is covered in whether n8n is worth learning for a solo freelance practice.

Testing before paying. The free self-hosted edition is a genuinely complete product. Building the first five workflows locally and only moving to Cloud once they earn something is the sensible order of operations.

When the Cloud plan pays for itself

Cloud earns its money the moment an automation stops being personal and starts being a client's.

The specific trigger is delivery risk. A workflow that sends a weekly digest can be down for a day and nothing happens. A workflow that routes a client's inbound leads cannot. Once an automation is billed on its outcome, the €24 stops being a cost and becomes the cheapest insurance in the stack — it buys uptime nobody has to personally watch.

The second trigger is plan-level features. Workflow history is the clearest one. ToolRadar's table, updated 4 September 2026 and not corroborated by any other guide gathered here, shows 1 day of history on Starter and 5 days on Pro. When a workflow fails on Thursday and a client reports it on Monday, one day of history is not enough to debug it. That difference between Starter and Pro is worth more to a working freelancer than the extra executions — assuming the underlying figures hold up when checked directly.

The third is concurrency. Starter is reported at 5 concurrent runs, Pro at 20 — again, from that single ToolRadar source. A webhook that can receive a burst of twenty submissions in a minute turns five concurrent slots into a queue, and a queue with a timeout behind it is a lost lead.

For anyone still deciding between n8n and the alternatives rather than between its hosting options, the head-to-head in n8n against Zapier for small business automation and the Make.com comparison cover that ground properly.

Three honest cons of paying for n8n Cloud

The execution ladder is steep in the middle. The jump from Pro at a reported €60 for 10,000 executions to Business at a reported €800 for 40,000 is enormous — four times the executions for more than thirteen times the price, as CloudZero's numbers work out. Anyone landing at 12,000 executions a month is in an awkward place, and the rational move is often to shift the high-volume workflow to a self-hosted instance and keep the client-facing ones on Cloud.

The plan limits are not all about executions, and they're not all confirmed. Active workflow counts, concurrency, shared projects and history all reportedly vary by tier. Arahi.ai, updated 10 July 2026, reports a 320 MiB RAM ceiling on Starter along with a 5-active-workflow limit — but Instapods, updated 9 September 2026, says every Cloud plan now includes unlimited workflows and unlimited users. Those two cannot both be true today, and n8n's own page doesn't resolve it in the sections checked for this article. A memory ceiling matters if large files or big JSON payloads are involved; check the official page for current limits rather than trusting any single summary, this one included.

Renting something that could be free. That sounds obvious and it still stings when the bill arrives during a slow month. The software itself is genuinely free to self host, and unlike most tools with an open core, the free version is not crippled. What Cloud is actually selling is operations; anyone with spare capacity to run their own operations is paying for a service they could render themselves.

Who should not pay for n8n Cloud

Anyone still learning the tool: run it on a laptop with Docker, break it, rebuild it, and come back to this question in two months with an actual execution volume in hand rather than a guess.

Anyone whose entire automation portfolio is personal — digests, backups, scrapers that feed a spreadsheet read on Fridays. None of that needs uptime.

Anyone with a sysadmin instinct and an existing server: the bill will be resented, and rightly so.

Anyone considering Business-tier prices without first checking whether a hybrid split works. Running the noisy, high-volume workflows on a self-hosted instance and the client-critical ones on Starter or Pro is a legitimate architecture, and it's roughly what the price gap between Pro and Business pushes toward.

A quick way to decide in five minutes

Count three numbers before looking at the pricing page again.

1. Executions per month. Add up every trigger honestly, including every scheduled poll. A five-minute schedule is 8,640 runs a month on its own.
2. The billable hour, times three. That's the rough monthly cost of self hosting for someone without existing infrastructure to piggyback on.
3. Client-facing workflows. Count how many automations, if down for six hours, would cost a client conversation nobody wants to have.

If number three is zero, self host. If number three is more than zero and number two is larger than €24, pay for the Cloud plan and stop thinking about it. That's genuinely the whole decision; everything else in this article is detail underneath it.

FAQ

Is n8n self hosted really free?

The licence is free. The self-hosted Community edition is reported by SyncGTM, in a review updated 26 August 2026, to cost nothing in licence fees with no stated cap on executions or workflows. What it is not is cost-free, because you pay for the server, the database and your own maintenance time. Budget a few dollars a month for a VPS and a few hours a month for keeping it healthy.

How many executions does the cheapest n8n Cloud plan include?

2,500 per month on the Starter plan, as several 2026 pricing guides report it, including ConnectSafely.ai on 10 September 2026 and CheckThat.ai, whose page carries a 13 September 2026 update date. An execution is one workflow run regardless of how many nodes it contains, which is what makes the number go further than it looks. Re-check the official pricing page before you commit, since the figure comes from third-party summaries.

Is annual billing on n8n Cloud worth it?

It is worth it if you are already confident you will use the tool for a year. CloudZero, on 10 September 2026, reports the annual discount at about 17%, and CheckThat.ai independently lists Starter at €20 a month billed annually against a €24 monthly rate. That saves roughly €48 a year on Starter, which is real money but not enough to justify locking in before you know your volume.

Can I move from self hosted to n8n Cloud later?

Yes, and this is the usual path. Workflows export as JSON and import into another instance, so the structure moves cleanly. The part that does not move automatically is credentials, because every OAuth connection has to be re-authorised on the new instance, so plan an afternoon for a portfolio of ten or so workflows.

What happens if I go over my execution limit?

The published tables do not make the overage behaviour clear, and the sources gathered for this article do not state it, so treat this as unknown and ask n8n directly before you build something that runs close to the ceiling. The safer design is to leave headroom. If you are consistently at 80% of your plan's executions, either upgrade or move one noisy workflow off Cloud.

Does the free self-hosted version have fewer features?

For a solo operator, effectively no. The features that separate the paid tiers are mostly organisational and enterprise-shaped: SSO, log streaming, external secrets and advanced permissions, which the 2026 guides list on the Business tier. The node library and the workflow engine are the same, which is why self hosting is a serious option rather than a trial.

The verdict

Pay for n8n Cloud when other people depend on the automations, self host when they do not. That is not a hedge, it is the actual line, and it sits somewhere around the first paying automation client. At a reported €24 a month for Starter, Cloud costs less than three billable hours a year at most freelance rates, and self hosting costs more than that in attention alone unless a server is already running for other reasons. Whoever should self host already knows it, because they're already running Postgres for something else.

The one thing worth doing before either choice is counting executions honestly, because that number decides the plan, and most people guess it wrong by a factor of three.

If inbox triage is the workflow driving this decision, an AI Email Triage Kit for n8n is available directly from this site for $19 — a set of importable workflows rather than a course.

Related reading

• Is n8n Worth Learning in 2026 for Solo Freelancers?
• n8n vs Zapier for Small Business Automation
• Make.com vs n8n: Which Automation Tool (2026)

About the author

The Tool Signal is a reference site on automation tooling for freelancers and small businesses. No figure in this article comes from a Cloud invoice or a self-hosted deployment held by the site; every price names the publisher and the date it was reported instead of claiming firsthand use. The hosting trade-offs described here follow from how n8n's execution model and self-hosting requirements are documented, not from a billing page, and the figures come from other people's reading of the vendor's pricing page, checked on 14 September 2026, cross-checked against each other where they overlapped and disagreed. Where a source left something unclear — overage behaviour, the Starter workflow limit, the exact Business annual price — this article says so rather than filling the gap. The full picture on n8n Cloud pricing is worth checking at the source before paying, because pricing pages change faster than articles do.