n8n pricing explained: cloud executions vs the real cost of self-hosting

How much does n8n cost?

n8n Cloud costs 20 EUR a month for 2,500 executions, 50 EUR a month for 10,000, and 667 EUR a month for 40,000, all on annual billing as listed on n8n.io/pricing on July 20, 2026. Enterprise is quoted. Self-hosting the Community Edition costs nothing in license fees. Monthly billing runs about 17% higher than the annual figures.

The plans differ on more than the execution count. Concurrency, the number of workflow runs that can be in flight at the same instant, is the limit teams hit second most often, and AI credits are metered separately from executions. Here is how the published tiers line up.

Plan	Price (annual billing)	Executions / month	Concurrency	Notable
Starter	20 EUR / mo	2,500	5	About 2,300 AI credits a month
Pro	50 EUR / mo	10,000	20	Up to 13,700 AI credits a month
Business	667 EUR / mo	40,000	Higher	Self-hosted option, 30 days of insights
Enterprise	Contact sales	Custom	200+	365 days of insights

US buyers should note the EUR listing. Your card gets charged a converted amount that moves with the exchange rate, so a plan quoted at 50 EUR is not a fixed dollar line on the budget. It is a small thing until finance asks why the number changed.

Is n8n free?

Yes, in one specific form. The n8n Community Edition is free and open source on GitHub, you can run it on your own server, and it has no execution cap at all. What you give up is the managed hosting, the enterprise features gated behind paid tiers, and any vendor responsibility for uptime. The Cloud plans are the paid product.

There is no free tier on Cloud, only a trial. So the real choice is not free versus paid, it is who runs the infrastructure. Teams with a platform engineer and an existing container setup often find self-hosting genuinely cheap. Teams without one end up paying for it in another currency, which we get into below.

Does n8n charge per step or per execution?

Per execution. One complete workflow run counts as one execution regardless of how many nodes or steps it contains, so a 40-node workflow with branches, loops, and error handling bills the same as a two-node one. This is the single biggest structural difference between n8n's pricing and per-task tools like Zapier, and it favors complex workflows heavily.

The practical effect: consolidate. If you can do the same job in one workflow with several branches instead of five workflows chained by webhooks, you cut your execution count by five. Sub-workflows called from a parent do count separately, so watch that pattern if you are optimizing. And since April 2026 n8n removed the active-workflow limits that used to cap how many workflows you could leave switched on, so there is no longer any reason to keep useful automations disabled to stay inside a tier.

What pushes your n8n bill up?

Four things, in roughly this order: high-frequency schedules, per-record loops that trigger child workflows, retries that re-run a whole workflow, and concurrency caps that force a tier jump before your execution count does. None of them are obvious when you build the first automation, and all of them compound quietly.

A schedule trigger firing every minute is 43,200 executions a month on its own, which overruns the 40,000 Business allowance with one workflow. Move it to every five minutes and it is 8,640. Most sync jobs do not need minute-level freshness, and the difference between those two settings is the difference between a 50 EUR plan and a 667 EUR one.

Retries are the sneaky one. If a workflow fails partway through and the retry re-runs the whole thing, an API that is flaky for an hour can double or triple the executions for that window. Concurrency is the other trap: Starter allows 5 concurrent runs and Pro allows 20, so a burst of webhooks can queue or fail at a volume well under your monthly execution allowance. You upgrade for the concurrency, not the executions.

Is self-hosting n8n worth it?

It depends on whether your team already runs servers. A small VPS capable of handling n8n for a modest workload costs a few dollars a month, so the infrastructure is close to free. The real cost is operational: version upgrades, database backups, TLS certificate renewal, queue mode setup at scale, monitoring, and someone reachable when a workflow silently stops at 2am.

Put a number on it honestly. If self-hosting costs you four hours a month of an engineer's time at a loaded rate of $90 an hour, that is $360 a month of hidden spend, which is seven times the Pro plan. If your team already has CI, container orchestration, and paging in place, the marginal hours are much closer to zero and self-hosting is the obviously correct call. The middle case, one technical person who inherits n8n alongside their real job, is where self-hosting quietly turns expensive.

If you do go self-hosted, treat it like production from day one. Pin the version, back up the Postgres database on a schedule you have actually tested restoring from, and put the instance behind a reverse proxy with automatic certificate renewal. Teams who do not want to handle server provisioning and zero-downtime deploys in-house should price that work in before they call the Community Edition free, because it is the part that keeps costing after launch.

Is n8n cheaper than Zapier?

For complex, multi-step workflows, usually yes, because n8n bills one execution per workflow run while Zapier bills per task, meaning per step. A 15-step workflow run 5,000 times a month is 5,000 executions on n8n and 75,000 tasks on Zapier. For simple two-step automations at low volume the gap narrows a lot.

The comparison that matters more for a data team is metered versus flat. Below is what the two shapes look like as volume climbs, using n8n's published EUR figures and our own published USD tiers. We are not going to pretend flat wins everywhere: at 10,000 runs a month, n8n Pro at 50 EUR is cheaper than anything we sell, and that is a real advantage of their model.

Monthly workflow runs	n8n Cloud plan needed	n8n price	Adapters flat tier
2,000	Starter	20 EUR / mo	Starter, $49 / mo
9,000	Pro	50 EUR / mo	Starter or Growth, $49 to $149 / mo
35,000	Business	667 EUR / mo	Growth, $149 / mo
120,000	Enterprise	Quoted by sales	Scale, $399 / mo

The interesting row is 35,000. That is where the execution meter stops being cheap, because the Pro to Business step multiplies your bill by more than thirteen while multiplying your allowance by four. High-frequency scheduled syncs are exactly the workload that walks into that step. If you want the arithmetic across the whole category, we work it through in what data integration costs, and the build-or-buy version of the same question is in build vs buy for integration.

Which one you should pick comes down to what you are actually doing. n8n is a general automation and AI workflow builder with hundreds of nodes and a code node when you need it, and for branching logic it is very good value. If your job is narrower, moving records between a database, a warehouse, and a few SaaS systems on a schedule with retries and logs, a purpose-built data integration platform gets you there with less to maintain and a bill that does not move with run count. Our n8n alternative page compares the two directly, the tiers are listed on pricing, and you can map a real table pair in the live demo in a couple of minutes.