n8n pricing 2026: Cloud plans (Starter, Pro, Business), self-hosted and the real cost

n8n pricing, checked on n8n.io/pricing on October 14, 2026: Starter €24 per month for 2,500 executions, Pro €60 per month for 10,000 executions, Business €800 per month for 40,000 executions (self-hosted only), Enterprise on quote; the self-hosted Community Edition is free with no execution quota. Billed annually, Starter comes to €20 per month, Pro to €50 and Business to €667 (n8n advertises a 17% saving). Amounts are the euro prices shown on the official page, which also offers contracts paid in US dollars; n8n revises them regularly, hence the date.

Plan	Monthly	Annual (per month)	Executions per month	Concurrent executions	Hosting
Community Edition	€0	€0	unlimited (bound = your server)	configurable (N8N_CONCURRENCY_PRODUCTION_LIMIT)	yours
Starter	€24	€20	2,500	5	n8n Cloud
Pro	€60	€50	10,000	20	n8n Cloud
Pro, 50,000 tier	€145	about €121 (€1,450 per year)	50,000	up to 50	n8n Cloud
Business	€800	€667	40,000 (tiers up to 500,000)	not stated	self-hosted only (Cloud on a waitlist)
Enterprise	on quote	on quote	custom	200+	Cloud or self-hosted

What the table does not tell you, and what weighs as much as the price: how n8n counts what you consume, and what you still pay outside the subscription (AI model calls and, when self-hosting, the server and the time to run it). The rest of this guide covers the billing model, what really separates the plans, how to estimate your volume, and when self-hosting becomes the rational option.

If you go with the free Community Edition, the FlowKit Self-Hosting Kit (€79, one-time payment) ships the production stack already tested: Docker Compose with PostgreSQL 18, HTTPS through Caddy, queue mode, verified encrypted backups, restore, and updates with rollback.

The key principle: n8n bills per execution, not per task

n8n Cloud's entire pricing logic fits in one definition: one execution = one full pass through the workflow, from the trigger to the last node. Whether a workflow contains 3 nodes or 60, whether it calls an AI model once or twelve times, whether it loops over 200 rows of a Google Sheet — that counts as a single execution against your monthly quota.

That is the exact opposite of the dominant model among competitors. Zapier counts every task (each action executed inside a Zap), Make counts every operation (each module traversed, multiplied by the number of items processed). On a simple two-step scenario, the difference is anecdotal. On a typical AI workflow — incoming webhook, enrichment, LLM call, parsing, CRM update, Slack notification, logging — a single trigger burns six or seven tasks on Zapier, one single execution on n8n. Multiply that by hundreds of triggers per day and the gap becomes structural, not marginal. This is the central argument of our n8n vs Make vs Zapier comparison, and the reason many teams end up migrating from Zapier to n8n once their automations get more complex.

The practical consequence: on n8n Cloud, making a workflow more sophisticated costs nothing extra. Adding a validation node, a retry step, a second AI call to double-check the output of the first — your bill does not move. This decoupling between workflow richness and cost is especially valuable for AI agents, which naturally chain many steps together.

Starter, Pro, Business, Enterprise: what actually varies between the plans

Every plan shares the same engine and the same nodes, with unlimited users and active workflows. What sets them apart is capacity and team features (checked on October 14, 2026):

• Monthly execution volume: 2,500 on Starter, 10,000 or 50,000 on Pro, 40,000 to 500,000 on Business. This is the quota to estimate before choosing (next section);
• Concurrent executions: 5 on Starter, 20 or 50 on Pro, 200+ on Enterprise. This ceiling becomes critical when webhooks arrive in bursts or slow AI workflows (30 to 60 seconds per run) stack up; beyond it, executions wait their turn;
• Maximum execution duration: 5 minutes on Starter, 40 minutes on Pro and Enterprise. A workflow that processes a long PDF or chains several AI calls can exceed 5 minutes: this ceiling, more than the quota, is often what pushes people to Pro;
• Saved execution history: 2,500 executions and 7 days on Starter, 25,000 and 30 days on Pro. Hitting the limit does not stop your workflows, only the oldest traces are removed;
• Built-in AI Assistant credits: 1,600 per month on Starter, 4,000 or 9,600 on Pro depending on the tier. They pay for the assistant that helps you build workflows, not for the AI calls your workflows make (see below);
• Team features: 1 shared project on Starter, 3 on Pro (with admin roles and workflow version history), 6 on Business with SSO/LDAP, multiple environments and Git version control (queue mode itself is already available in the free Community Edition); Enterprise adds an external secret store, log streaming, extended retention and SLA-backed support.

Two points often surprise people. First, Business is not available on n8n Cloud at this date: it is a license key you apply to your own instance, which pings n8n's license server daily and reports the number of production executions. Second, n8n advertises 50% off Business for companies under 20 employees that qualify for its Start-up plan. The n8n Cloud free trial, no credit card required, gives Pro-level features with 1,000 executions, 5 concurrent executions and a 180-second maximum per execution.

Estimating your real execution volume before subscribing

The classic mistake is picking a plan on gut feeling, then discovering a blown quota at the end of the month — or, conversely, an oversized plan. The method is simple: list your planned workflows and, for each one, work out the number of monthly triggers.

• Scheduled workflows (Schedule Trigger): the math is deterministic. A workflow running every 15 minutes means 4 × 24 × 30 ≈ 2,880 executions per month — even if it finds nothing to process most of the time. A thrifty reflex: stretch the interval of polls that usually come back empty, or replace a frequent poll with a webhook whenever the source application allows it;
• Event-driven workflows (Webhook, Chat Trigger, inbound email): volume follows your actual activity. Base it on your existing metrics (orders/month, leads/month, messages/month) rather than intuition;
• Sub-workflows: a workflow called by another via Execute Workflow counts too — a pipeline heavily split into sub-workflows consumes more executions than an equivalent monolith, something to weigh in your architecture;
• Error margin: add 20 to 30% for retries, testing and growth, then monitor actual consumption during the first month in the dashboard.

What the subscription does not cover: your AI API calls

A point often discovered too late: n8n Cloud bills the orchestration, not the intelligence. Every call to OpenAI, Anthropic, Mistral or any other provider goes through your own API keys and is billed to you separately, per token. An AI agent workflow running 1,000 times a month can therefore cost more in tokens than in n8n subscription — especially if the prompt carries a long context on every turn. Budget the two lines separately, and equip yourself to track the second one: our guide to tracking the cost of AI calls in n8n shows how to log token consumption workflow by workflow, before the provider's invoice becomes your alerting system.

When self-hosting becomes the rational option

Since n8n is source-available, the alternative to Cloud is hosting it yourself — with executions then unlimited, within the capacity of your server (the concurrency, payload and license limits that do apply are listed in our guide Is n8n free?). The switch becomes rational in three situations: high, predictable volume (tens of thousands of monthly executions make the cloud quota structurally uncomfortable), sensitive data that must not transit through third-party infrastructure (healthcare, legal, HR), or a need for fine-grained control (pinned version, community nodes, memory tuning, queue mode with Redis to scale AI workflows).

But self-hosting is not free either: it shifts the cost from the subscription to the server and, above all, to human time — installation, updates, backups, monitoring, incident recovery. That hidden cost is systematically underestimated; we quantified it in detail in our analysis of the real cost of self-hosted n8n, and the full self-hosted vs Cloud comparison helps you decide based on your profile.

This tension between renting and owning is nothing new. The landmark paper by Armbrust and co-authors (Berkeley), "A View of Cloud Computing", published in 2010 in Communications of the ACM (see the study), formalizes what the cloud actually sells: elasticity (paying as you go rather than provisioning for the peak) and the transfer of the risk of over- or under-provisioning to the provider. Applied to n8n: if your execution volume is irregular or growing unpredictably, the Cloud absorbs that risk for you; if it is massive and stable, you are paying an elasticity premium you no longer need — and owning the infrastructure makes sense again.

A five-question decision checklist

1. How many executions per month? Do the math trigger by trigger (section above), not by instinct;
2. Are your AI workflows slow? If so, look at the concurrent execution ceiling as much as the monthly quota;
3. Who administers it? Without available DevOps skills, the Cloud premium is almost always lower than the cost of administration time;
4. Can your data leave your infrastructure? If not, the cloud plan question does not even arise: self-hosted;
5. Is your volume predictable? Irregular or fast-growing → Cloud (elasticity works in your favor); massive and stable → self-hosting is probably more rational.

And in every case: the amounts in this guide date from October 14, 2026, so double-check them on n8n.io/pricing before committing.

Making every execution pay off from month one

Whichever plan you pick, the real economic lever is not the pricing grid: it is the time your workflows save once they are running. Rather than burning your first weeks of subscription building from an empty canvas, our catalog of ready-to-use n8n workflows provides pipelines already tested — AI agents, RAG, CRM and e-commerce automations — to import and adapt with a few settings, so that every execution deducted from your quota produces value from day one.