n8n Cloud pricing in 2026, explained, and when self-hosting is cheaper

I ran n8n both ways. For a quarter it lived on a small server I rented, with a database next to it and a domain pointing at it. Then I moved the same workflows to n8n Cloud, because I was tired of being the person who ran the updates on a Sunday night. Both work. They cost different things, and the pricing page only shows one of them.

This page is what n8n counts, what each plan costs, and when running it yourself is cheaper.

An execution is the unit

n8n bills by the execution. An execution is one run of a workflow, from the trigger to the last node. Three nodes or thirty, it counts as one, and n8n does not limit the steps in an execution.

The AI is not included. You add your own OpenAI or Anthropic key to the node, and the provider bills the tokens. So an n8n month is two bills.

The plans

Plan	Price	Executions a month	Past the limit
Community Edition	Free, self-hosted	Set by your server	Not billed
Starter	€20 a month billed annually, or $24 billed monthly	2,500	Capped, no overages
Pro	€50 a month billed annually, or $60 billed monthly	10,000	Capped, no overages
Business	See n8n's pricing page	Set by the plan	Keeps running, €4,000 or $4,800 per extra 300,000

The pricing page shows euros on annual billing and switches to dollars on the monthly toggle, so each plan has two numbers.

A five-person team's month

These are the three automations a small team sets up first, at the same volumes I use on what an AI automation costs per run: inbox triage on 300 emails, a daily report on 22 working days, and a prep brief for 60 meetings. That is 382 executions, well inside Starter's 2,500.

The AI bill depends on the model. I use three illustrative tiers per million tokens in and out: small at $0.15 and $0.60, mid at $1 and $5, large at $3 and $15.

Automation	Runs	Tokens per run	Small	Mid	Large
Inbox triage	300	6,000 in, 800 out	$0.41	$3.00	$9.00
Daily report	22	20,000 in, 1,500 out	$0.09	$0.61	$1.82
Meeting prep	60	10,000 in, 600 out	$0.11	$0.78	$2.34
AI bill for the month			$0.61	$4.39	$13.16

To check one line: inbox triage on the mid tier is 6,000 tokens at $1 per million, $0.006, plus 800 at $5 per million, $0.004. That is $0.01 a run, and $3.00 for 300 runs.

So the month on n8n Cloud is Starter at €20 billed annually, or $24 monthly, plus $0.61 to $13 from your provider. That is a fair price, nowhere near the cap.

What the cap means

n8n's pricing page says Starter and Pro are "capped and have no overages". No surprise invoice, and once the month's executions are spent, nothing more runs that month.

I once built a retry that rescheduled itself every minute after a failed API call. It failed all night, each retry an execution. A loop like that can spend 2,500 executions in a night, and the next morning's inbox triage does not run.

Only Business keeps running past its limit, at €4,000 or $4,800 per extra 300,000 executions. On Starter and Pro the guard is your own: a retry limit on every workflow.

When self-hosting is cheaper

The Community Edition is free under the Sustainable Use License, for internal business or non-commercial use, which covers most small teams automating their own work.

The real costs are a server, updates, backups, a database, a domain, and the person who owns all of it.

Self-hosting is cheaper in money from day one and more expensive in someone's hours. When I self-hosted, the Sunday updates cost more than the server. If a developer who likes the work does them, self-hosting wins. If it is the founder, Cloud is cheaper even at €20 a month.

Who should stay on n8n

Stay on n8n if someone on your team is technical and happy to own it. Nothing else is as flexible for the price. If your workflows are data pipelines with code nodes and custom HTTP calls, n8n does that better than tools built for people who do not code. It connects to hundreds of integrations, and thousands of community templates mean someone has usually built the job before you.

The same month on Servflow

Servflow counts differently. Each automation runs as one agent, and a run costs the AI at cost plus a fixed fee. Credits pay for the AI at what it costs us, plus 4 credits a run. 1 credit is $0.001. Free is 4,000 credits once, no card, and the plans are $10, $25 and $50 a month for 10,000, 26,000 and 55,000 credits.

The same 382 runs, with the AI included:

Tier	Credits per run (triage, report, prep)	Month in credits	Plan that covers it
Small	6, 9, 6	2,358	$10
Mid	15, 33, 18	6,306	$10
Large	36, 92, 46	15,584	$25

To be straight: n8n Starter covers these executions, and the AI bill on top is small. What you buy on Servflow is different. No server to keep, no canvas to wire, a template to start from, the AI and the run on one number, and the cost shown on every run. If you would rather keep your provider, use your own key and pay only the 4-credit run fee.

The cap works the other way round too. Credits stop at zero. Nothing is billed past your plan unless you top up, at $1 per 1,000 credits. Servflow is also SaaS or self-hosted, the same product, so the choice in the section above is yours here as well.

One run opened in the Servflow dashboard: the agent name with a succeeded badge, Answered in 36 s, a What happened list with the schedule trigger, the sub agent as step 1 and each tool it called as steps 2 to 8, and a Details panel showing Started, Took 36 s, Credits used 4, Trigger Schedule — One run on Servflow, with its credit cost in the Details panel. This one ran on the team's own key, so it shows the 4-credit run fee alone.

How this runs on Servflow

Every run is a trace: the trigger, each step, each model call with its tokens, and the credits it used. The runs list has a Credits column, and the month's bill is its sum. n8n alternatives covers moving a workflow across.

FAQ

How much does n8n Cloud cost? Starter is €20 a month billed annually, or $24 a month billed monthly, for 2,500 workflow executions. Pro is €50 a month billed annually, or $60 a month billed monthly, for 10,000. Business keeps running past its limit at €4,000 or $4,800 per extra 300,000 executions.

What counts as an execution in n8n? One run of a workflow from its trigger to its end. A workflow with three nodes and one with thirty both count as one execution, and n8n does not limit the steps in each. Calls to an AI model inside the workflow are billed separately by the model provider.

What happens when I run out of n8n Cloud executions? On Starter and Pro, n8n's pricing page says the plans are capped and have no overages, so once the month's executions are spent nothing more runs that month. Only Business keeps running past its limit, at €4,000 or $4,800 per extra 300,000 executions.

Is self-hosting n8n cheaper than n8n Cloud? In money, yes, from the first day. The Community Edition is free under the Sustainable Use License for internal business or non-commercial use. In time, no: someone owns the server, the updates, the backups, the database and the domain. It is cheaper when that person already exists and enjoys the work.

Does n8n Cloud include the AI? No. You add your own OpenAI or Anthropic key to the AI node and the provider bills you for the tokens. For inbox triage, a daily report and meeting prep at a small team's volume, that is under $1 to about $13 a month by model tier.

Start free

4,000 credits, no card, and every run shows what it cost. From $10 a month, AI at cost.

Start free