n8n pricing 2026: what each workflow run costs

You will leave this n8n pricing guide with the right plan and a monthly run estimate for your busiest automation. Set aside 30 minutes with n8n's Executions screen and a calculator. The important choice is not 20 euros versus 50 euros. It is whether your workflow volume, simultaneous jobs, and maintenance burden fit the plan.

Pricing was checked August 8, 2026, against the official n8n pricing page and execution documentation. Listed prices exclude sales tax and use annual billing.

n8n pricing at a glance

The annual commitment works out to €240 for Starter, €600 for Pro, and €8,004 for Business before tax. Business is not simply a larger hosted plan. It runs on infrastructure you control and pays for management features such as separate development and production environments, version control, and company login controls.

n8n is automation software. It moves information between the systems you already use, without someone copying each record by hand. A workflow is the full routine. A node is one box inside it, such as receiving a form, checking a phone number, updating the software that stores your customers, and sending an alert.

What n8n charges for

n8n calls one complete run an execution. On paid plans, only production executions count toward the allowance. Production means a published workflow starts automatically from a schedule, an incoming web address, or a check for fresh data.

The number of nodes does not change the execution count. If one new inquiry starts a 12-node workflow, the run normally uses one execution. That billing shape favors longer routines because the extra boxes do not each create another charge.

The trigger still matters:

A webhook is a web address that catches data another system sends. It is often the cleaner choice for lead capture because the workflow starts only when a form actually arrives. A schedule that runs every five minutes starts about 8,640 times in a 30-day month. That single timer cannot fit inside Starter's 2,500-run allowance.

The real bill at three volumes

The smallest useful estimate starts with business events, not nodes. Count form submissions, orders, booked appointments, report schedules, and chat messages that cause a workflow to start.

Suppose you run three routines:

That workload fits Starter with 1,040 runs left for growth, retries, and seasonal variation. The number of steps inside each routine does not alter that total.

At 3,000 runs, Starter is too small. Pro provides 10,000. At 9,000, Pro technically fits, but an 11% rise would cross the allowance. Do not run a customer-facing system against the ceiling. Leave at least one busy month's normal variation between your forecast and the plan limit.

Business changes the question. Its €667 monthly price includes 40,000 executions, but n8n still runs on your server. A company buys this tier for controls, not cheap hosting: different environments for testing and live work, version history stored with Git (software that records what changed), and SSO, a single company login that can remove access centrally.

Starter versus Pro

Choose Starter when one person owns the workflows, 2,500 monthly production runs are enough, and no more than five executions need to happen at the same time. All paid plans include unlimited users, workflows, steps, and integrations. Starter is not capped at a tiny number of automations.

Choose Pro when volume crosses 2,500 or the business needs more working room. It raises the published allowance to 10,000 executions and the concurrency limit to 20. Concurrency means how many jobs can run at the same moment. A burst of website forms can queue if the limit is already occupied by slow jobs.

Pro also adds three shared projects instead of one, admin roles, global variables, workflow history, and execution search. A global variable is one saved value, such as a reporting email address, that several workflows can reuse. These controls matter when more than one person changes automations.

The price jump is €30 a month on annual billing. If you only need more runs, divide that difference by the extra 7,500 executions. The capacity costs €0.004 per additional available run, but that figure is useful only if you will use the room. Paying for unused headroom is still wasted money.

For a broader platform decision, read n8n vs Make. That comparison explains why n8n's full-run meter can beat per-action billing on long workflows. The automation tools cost map shows the other billing models beside it.

Cloud versus Community edition

n8n Cloud means n8n runs the servers. You build the workflows and n8n handles the underlying service. For most owners, that transfers a recurring technical job out of the business.

Community edition removes the software subscription. Self-hosting means the application runs on a server you control. According to n8n's hosting documentation, you can install it with Docker, Kubernetes, npm, or a cloud provider. Those are different technical ways to package and operate the software.

The word “free” covers the license, not the operation. Someone must install security updates, back up the database, renew the web certificate, monitor storage, protect saved passwords, and restore the system after a failure. If nobody is explicitly responsible for those tasks, Community edition has an unknown labor bill and a weak recovery plan.

Use Community when you have a capable technical owner, need control over where data is stored, and can prove that backups restore correctly. Use Cloud when the automation must stay available but server work is outside your business.

Do not move solely to save €20 a month. One missed lead or one hour of emergency repair can erase several months of that difference.

The part that breaks

The first trap is a timer that runs too often. A scheduled workflow consumes a production execution whenever it fires, including a run that finds no useful work. Open each Schedule Trigger node and ask how quickly the business really needs the result. A daily stock report does not need a five-minute schedule.

The second trap is confusing run allowance with retained history. n8n deletes older saved execution details when a plan's storage or retention boundary is reached. That does not stop new workflows from running, but it can remove the evidence you need to diagnose an old failure. Export important business records to their system of record instead of treating execution history as permanent storage.

The third trap is ignoring concurrency. Ten leads arriving together are ten jobs, even if the monthly total is modest. Test a burst before launch. Slow AI steps and large data jobs can occupy all available slots while customer work waits.

Finally, a failed external service can trigger repeated attempts. Check Executions after adding retry logic. A helpful retry policy can become a volume problem if an app remains unavailable for hours.

Copy this

Use this sheet for every production workflow:

workflow name:
trigger type: webhook / polling / schedule
events per business day:
business days per month:
fixed scheduled runs per month:

estimated monthly executions =
(events per business day × business days per month)
+ fixed scheduled runs per month

largest likely one-hour burst:
longest normal run time:
named owner:
backup owner:
failure alert destination:

Add the estimates, then multiply the total by 1.25 for a 25% operating cushion. Compare that number with 2,500 for Starter and 10,000 for Pro. Check the largest one-hour burst against the plan's concurrency, not its monthly allowance.

Upgrade path

Start by removing waste, then buy capacity. Replace unnecessary schedules with webhooks, slow the checks that do not need immediate answers, and keep manual testing unpublished so those runs stay outside the quota.

When Starter no longer fits, move to Pro before the busy month begins. When Pro's collaboration controls are still too loose, document why you need Business or Enterprise before accepting the self-hosting work or a custom contract.

If the meter itself is wrong for your workflow, compare the Make pricing model and Zapier pricing structure. Then use the automation runbooks to build the next routine on the platform your team can keep working.