n8n Pricing Explained in 2026: Self-Hosted vs Cloud Costs

Ask three people what n8n costs and you will hear three different answers, and all three will be right. n8n is the rare automation tool with two completely separate price tracks: a hosted cloud plan that bills per workflow execution, and a free self-hosted version with no execution limits at all. In 2026, that means one small business pays €20 a month while a similar one pays nothing but the price of a £7 VPS.

This guide lays out every 2026 plan, explains the execution billing model that makes n8n so different from Zapier and Make, and puts real UK self-hosting numbers next to the cloud prices so you can decide which track fits your business.

n8n's 2026 plans at a glance

n8n is headquartered in Berlin and bills in euros. Cloud data for European customers lives in Frankfurt. Prices below are per month, with annual billing listed first since it carries roughly a 17 percent discount.

Two details apply across every paid plan: users and workflows are unlimited. You are not counting seats or active workflows anymore. The meter that matters is executions.

The billing unit that changes everything

An n8n execution is one complete run of a workflow, no matter how many steps it has. A workflow with 25 nodes that syncs your CRM, posts to Slack, and updates a spreadsheet counts as exactly one execution.

This is the opposite of how most automation tools bill. On Zapier, every completed action step is a task, so a 25-step run can cost roughly 25 tasks. On Make, every module step is an operation, so the same run costs about 25 operations. For simple two-step workflows this barely matters. For the multi-step agent and AI workflows businesses are building in 2026, it is the whole pricing story: n8n gets cheaper per workload exactly when the workflow gets more complex. Our comparison of n8n, Zapier, and Make head to head shows the same structural difference from the vendor-choice angle.

There is one more meter to know about. Concurrent executions cap how many workflow runs can execute at the same instant: 5 on Starter, 20 on Pro, 30 on Business. Hit the cap and new runs queue rather than fail. It governs speed, not your monthly execution total.

Cloud plan details

Starter (€20/mo annual). 2,500 executions, one shared project, 2,300 AI Assistant credits. A shared project is n8n's container for grouping workflows and credentials so a small team collaborates under one set of permissions. This plan suits a business testing n8n in production or running a handful of low-frequency workflows.

Pro (€50/mo annual). 10,000 executions, admin roles, workflow history, up to 13,700 AI credits, and up to three shared projects. The default paid plan for most small production teams, and the plan most businesses land on once they depend on their automations.

Business (€667/mo annual). 40,000 executions, SSO, Git version control, environments, self-hosted only. This is the bridge tier for companies under roughly 100 employees that need collaboration and scale but want their data on their own infrastructure.

Enterprise (custom). SSO, audit logs, log streaming, dedicated support SLA, invoice billing, and air-gap deployment options. The plan for regulated teams.

Self-hosted Community: genuinely free, with a real cost

The self-hosted Community Edition has no execution meter, no workflow limit, and no user limit. Your only bill is the server. For a UK business in 2026, that looks like this:

• Trying it out: a £4 to £8 a month VPS (2 vCPU, 4 GB RAM, 50 GB SSD) handles light testing. 4 GB of RAM is the practical minimum for the current Docker Compose setup.
• Running a small business on it: £10 to £20 a month buys 2 to 4 vCPUs with 4 to 8 GB of RAM and NVMe storage, enough for real daily workflows with headroom. RAM is n8n's real bottleneck, because each active execution holds its payload in memory, and one large webhook body can eat 150 to 300 MB on its own.
• Production with backups and monitoring: £20 to £40 a month for a bigger box plus snapshots and offsite backup copies.

The hidden cost is you. Self-hosting needs someone comfortable with Linux, Docker, firewall rules, SSL certificates (free via Let's Encrypt), and a monthly patch cycle. Security, backups, and scaling are yours. For a technical team, that is a weekend project. For a business with no technical owner, it is a bill in disguise: either you learn the server, or you pay someone to babysit it. A good rule of thumb: if no one on your team has ever run a Docker Compose file, price in either a managed host or a cloud plan.

How many executions will you actually use?

This is where people blow the budget, because execution math is unintuitive. A few worked examples:

A workflow that runs every 5 minutes fires 288 times a day, or about 8,640 executions a month. That alone exceeds Starter's 2,500 and nearly fills Pro's 10,000. Polling workflows are the number one cause of surprise execution bills.

A 25-node lead-routing workflow that runs 200 times a month costs 200 executions on n8n cloud. On a per-step tool, the same runs cost roughly 5,000 billable units. Complexity is where n8n wins.

AI agent workflows are the new budget risk. Agent loops that call tools repeatedly can spin executions fast, and Starter's 2,300 AI Assistant credits (13,700 on Pro) are a separate meter from executions. Budget AI-heavy workflows with headroom, and measure a week of real usage before annual billing.

What the same workload costs across the four tracks

Take a business running 1,000 multi-step workflow runs a month, averaging 15 steps per run, with five team members:

Double the runs to 2,000 a month and Starter is dead (quota exceeded). At 50,000 runs, self-hosting is still the £10 to £20 VPS, while cloud Pro is two tiers short. The crossover is consistent: self-hosting is cheapest at any volume, but only if someone owns the server.

Euro prices above are converted at roughly €1.17 to the pound, which moves. Most international automation vendors bill UK customers in euros or dollars, so confirm the currency at checkout and ask your accountant how VAT applies to the invoice before you model the annual cost. For context on total project budgets, see what AI automation typically costs a small business.

Three UK businesses, three right answers

A Leeds estate agency, 12 staff, no technical team. 300 workflow runs a month, mostly lead alerts and viewing follow-ups. They pay €50 a month for Pro. Self-hosting would be cheaper in cash and a disaster in practice, because nobody there has ever SSH'd into a server. The best plan is the one someone can actually run.

A Manchester e-commerce store, 2 developers on staff. 4,000 workflow runs a month across order processing, stock alerts, and review requests. They self-host Community on a £14 a month VPS. Their developers already run Docker for the store, so n8n is one more Compose file. Annual cost: about £170, with effectively unlimited executions.

A London consultancy, 40 staff, compliance-conscious. Data must stay on company infrastructure, workflows need version control and SSO. Business at €667 a month, self-hosted. It is the priciest published tier, and the right one, because compliance, not cash, is the constraint.

The right n8n setup is the one your team can actually maintain, at the execution volume you actually run. Tech Instance builds and hosts production n8n workflows for businesses that want the automation without the server admin: CRM syncs, lead routing, review requests, AI agents, and the unglamorous plumbing in between. If you are still deciding what to automate first, see what an AI automation agency actually does. Browse our services, check transparent pricing, or get in touch for a quote scoped to your real workload.

Book a Strategy Call