n8n Review 2026: Per-Execution Pricing, Self-Hosting, AI Agents and the 3.0 Upgrade

Verdict at a glance

n8n is the strongest choice here for technical teams that build long, multi-step or AI-heavy workflows, because a 30-step run costs the same as a 3-step run. Self-hosting is free for internal use and has no execution caps. The trade-offs are a steeper learning curve than Zapier, euro-only pricing, non-refundable fees, and a Business plan you cannot get on n8n Cloud.

What works

• Per-execution billing: unlimited steps per run on every plan.
• Unlimited users, unlimited active workflows and every integration on all plans.
• Free, self-hosted Community Edition with no execution caps.
• Strong AI tooling: a LangChain-based AI Agent node, MCP server and client nodes, and an AI Workflow Builder.
• Cloud data stored in the EU, in Frankfurt.

What to watch

• Prices are shown in euros only.
• Business (667€ a month, billed annually) is self-hosted only.
• Starter caps a single execution at 5 minutes.
• All fees are non-refundable, and a lapsed Cloud trial instance is deleted.
• The Community Edition is fair-code, not OSI open source, and resale needs a commercial agreement.
• n8n 3.0 is a one-way upgrade, and it drops npm installs for self-hosters.

What n8n is in 2026

n8n is a visual workflow builder. You drag nodes onto a canvas, connect a trigger (a new email, a webhook, a schedule) to actions in other apps, and add logic, code and AI steps in between. Its integrations page advertises 2,000+ integrations, and a Code node covers anything that is not built in.

Two things set it apart from Zapier and Make. First, you can host it yourself, on your own server, with your data never leaving it. Second, it leans hard into developer features: JavaScript and Python code nodes, Git version control on higher plans, queue mode for scaling, and custom nodes on self-hosted installs.

The project is large and active. The GitHub repository had 209,381 stars and 61,359 forks on 3 November 2026. In October 2025 n8n raised a $180M Series C led by Accel at a $2.5bn valuation, and in May 2026 SAP bought shares in a secondary sale that valued n8n at $5.2bn. That deal also brings n8n into SAP's Joule Studio. At its first conference, In The Loop in Berlin on 13-14 October 2026, n8n launched standalone Agents and, according to an attendee's recap of the keynote, claimed 6.7 million active workflows a month. A light rebrand with a new logo landed the same week.

n8n pricing in November 2026

Every plan includes unlimited users, unlimited workflows and every integration. What changes between plans is the monthly execution quota, concurrency, log retention and admin features. n8n defines an execution as "a single run of your entire workflow. It doesn't matter how many steps are in the workflow or how much data it processes."

Official prices are in euros only. The headline figures are billed annually. Paying month to month costs roughly 20% more, which is the "Save 17%" the annual toggle advertises.

• Starter (n8n Cloud): 20€ a month billed annually, or about 24€ month to month. 2,500 executions, 5 concurrent executions, 1 shared project, 1,600 AI Assistant credits a month, forum support.
• Pro (n8n Cloud): 50€ a month billed annually for 10,000 executions, or about 60€ month to month. A 50,000-execution tier costs about 145€ month to month. Adds 3 shared projects, up to 50 concurrent executions, admin roles, workflow history, 7 days of insights and up to 9,600 Assistant credits.
• Business (self-hosted only): 667€ a month billed annually for 40,000 executions, or about 800€ month to month. Higher tiers start at about 985€ a month for 50,000 executions and run to about 5,245€ for 500,000. Adds SSO and LDAP, multiple environments, Git version control, queue mode, 6 shared projects and 30 days of insights.
• Enterprise (Cloud or self-hosted): custom pricing. Adds unlimited shared projects, 200+ concurrent executions, 365 days of insights, an external secret store, log streaming and an SLA.
• Community Edition (self-hosted): free, with no execution caps, but without the team and governance features listed below.

Startups get 50% off Business: 333€ a month billed annually for 40,000 executions. You qualify with fewer than 20 employees and less than €5M in total funding. The discount lasts a year, or longer while you stay eligible.

Per-execution billing vs Zapier and Make

This is the main reason people switch. Zapier counts a task for each action step that runs, and Make charges credits for each module run. On both, a long workflow multiplies your usage. On n8n, a workflow run is one execution whether it has 3 nodes or 40, and n8n says that "if it doesn't run, you don't pay."

Here is an example. Take a lead-routing workflow with 12 steps: enrich, score, branch, write to a CRM, post to Slack and so on. If it runs 2,000 times a month, that is 2,000 executions on n8n, which fits Pro's 10,000 quota with room to spare. On a per-step tool, the same volume is tens of thousands of billable steps.

The reverse also holds. If your automations are short, two-step jobs that run very often, the per-step model loses most of its penalty, and Zapier's larger app catalogue and easier editor may matter more. Some third-party guides say manual test runs and retries also count as executions. n8n's pricing page does not spell this out, so budget some headroom.

• Zapier Review 2026: Pricing, Task Limits, AI Costs and the Fine Print ↗
• Make.com Review 2026: New Prices, Credits, AI Agents and the Fine Print ↗
• Zapier vs Make (Integromat): Which Automation Platform Wins? ↗

Cloud plan limits that actually bite

Execution quotas get the attention, but the comparison table on n8n's pricing page hides limits that matter more for some workloads. For Starter, Pro and Enterprise respectively:

• Max execution duration: 5 minutes, 40 minutes, 40 minutes. A long scraping or AI batch job will not fit in Starter.
• Max saved executions: 2,500, 25,000 and 50,000.
• Execution log retention: 7 days, 30 days and unlimited.
• Workflow history: 1 day, 5 days and 365+ days.
• Concurrent executions: 5 on Starter, 20 on the 10,000-execution Pro tier and 50 on the higher Pro tier.
• Overages: capped on Starter and Pro, custom on Enterprise.
• Custom nodes and bash scripts are available on self-hosted installs only.

"Capped" means you cannot buy extra usage on that plan. Once you hit the quota you have to move up a tier. On Business and Enterprise, workflows keep running past the quota, but overage charges may apply if you do not upgrade.

Self-hosting: the free Community Edition

The Community Edition is free, self-hosted and has no execution caps, which is why many teams start there. You pay for your own server and your own maintenance, but not per run. A free Registered Community licence, which you get by entering your email, also unlocks folders, debug in editor and custom execution data.

What the Community Edition leaves out is mostly team and governance tooling: projects, sharing of workflows and credentials, SSO (SAML and LDAP), environments, Git version control, custom variables, external secrets, external binary data storage, log streaming and multi-main mode. Queue mode is included. For a solo builder this rarely matters. For a team of ten sharing credentials, it is the reason the Business plan exists.

AI agents, MCP and the Assistant

AI is where n8n has invested most. The AI Agent node is built on LangChain and can be wired to a chat model, memory and tools, including other workflows. Every plan includes the AI Agent node, the MCP Server Trigger, the MCP Client node and an instance-level MCP server.

In practice, MCP works both ways. The MCP Server Trigger lets an outside AI client call your n8n workflows as tools, and the MCP Client node lets an n8n agent use tools from other MCP servers. The instance-level MCP server lets AI coding tools such as Claude build workflows in your instance directly.

The AI Workflow Builder and n8n Assistant let you create, refine and debug workflows in plain language. On Cloud they run on monthly credits (1,600 on Starter, 4,000 or 9,600 on Pro, depending on tier) that do not roll over. Cloud plans can also use AI models without your own API keys. On self-hosted you bring your own key, so no credit limits apply.

n8n 3.0: what changes and who should wait

n8n 3.0.0 entered early-access beta on 27 October 2026, and the stable release is due in early December. It covers self-hosted, Cloud and Enterprise installs. On Cloud you opt in by choosing "3.0.0 Beta (early access)". Credentials and execution history carry over, and no pricing or licence changes came with it.

The breaking changes that matter most:

• Self-hosted installs require Docker. Running with npm or npx n8n is no longer supported, and n8n offers a one-line setup script instead. Kubernetes and Helm setups are unaffected.
• Legacy nodes are removed, including Function, Function Item, Item Lists, the legacy OpenAI node and older AI Agent modes.
• Cron and Interval are replaced by the Schedule Trigger.
• The Code node's default timeout drops from 5 minutes to 1 minute.
• Manual executions always run on workers, and the in-memory binary data mode is gone.
• Encryption key rotation is always on, and Slack credentials need a signing secret.
• Unverified community packages and Chat Hub are off by default.

Support timelines set the pace. 1.x support ends on 31 December 2026. 2.x gets security and critical fixes for 12 months after 3.0 goes stable, but new features ship on 3.x only. The Migration Report, under Settings, needs version 2.39 or later. If you run production on npm, plan the Docker move now rather than in December.

Security, data and the fine print

• Data location: hosted plans store data in the EU, on Microsoft Azure servers in Frankfurt, Germany.
• Security: n8n says it aligns its security program to SOC 2, with annual audits by an independent auditor. Enterprise customers can get the SOC 2 report, and everyone else the SOC 3 report.
• Refunds: all fees on monthly and annual plans are non-refundable, and n8n does not offer pro-rata refunds. EU consumers may still have a statutory withdrawal right, with exceptions.
• Cancellation: monthly plans can be cancelled at any time and run to the end of the month. Annual plans end at the end of the term. All plans auto-renew.
• Price changes: n8n can change prices and overage rates after giving "sufficient notice".
• Trial: if you do not upgrade, the trial expires and n8n deletes your instance. You then have 90 days to download your workflows.

n8n's self-serve terms were last updated on 26 October 2026. Read them before you commit to an annual plan.

The free trial

n8n Cloud offers a 14-day trial of Pro features with no credit card. It is tighter than Pro itself: 1,000 executions, 5 concurrent executions, a 180-second execution timeout, 1 team project, 800 Assistant credits and 100 AI credits. That is enough to build and test a few real workflows, but not to load-test them. The Business trial also lasts 14 days and does need a card.

Who should choose n8n

• Choose n8n Cloud Pro if you run long or AI-heavy workflows at moderate volume and want n8n to handle hosting.
• Choose the self-hosted Community Edition if you have someone comfortable with Docker, want unlimited runs, or need data kept on your own infrastructure.
• Choose Business or Enterprise if you need SSO, Git-based environments and governance. Business means hosting it yourself.
• Look elsewhere if your team is non-technical and your automations are short and simple. Zapier's editor is gentler, and per-step billing hurts less there.

Frequently asked questions

How much does n8n cost?

n8n Cloud starts at 20€ a month billed annually (about 24€ month to month) for 2,500 executions. Pro is 50€ a month billed annually for 10,000 executions, and the self-hosted Business plan is 667€ a month billed annually for 40,000. The self-hosted Community Edition is free.

Is n8n free?

Yes, if you host it yourself. The Community Edition has no execution caps but lacks team features such as projects, sharing, SSO and Git. There is no permanent free plan on n8n Cloud, only a 14-day trial.

Does n8n charge per step like Zapier?

No. n8n charges per workflow execution, and one run of the whole workflow counts once however many steps it has. Zapier bills per task and Make per credit, so long workflows cost more there.

Is n8n open source?

Not in the OSI sense. It is fair-code under the Sustainable Use License: free to use and modify for internal business, personal or non-commercial purposes, while resale or hosting it as a paid service needs a commercial agreement.

Can I get the Business plan on n8n Cloud?

Not yet. Business is self-hosted only. On n8n Cloud, SSO, environments and Git version control come with Enterprise.

What happens if I go over my execution quota?

On Starter and Pro the quota is capped, so you have to upgrade to the next tier to keep going. On Business and Enterprise, workflows keep running and overage charges may apply if you do not upgrade.

Does n8n offer refunds?

No. Fees on both monthly and annual plans are non-refundable, with no pro-rata refunds. You can cancel a monthly plan at any time and keep access until the end of that month.

Where does n8n Cloud store my data?

In the EU. Hosted plans store data on Microsoft Azure servers in Frankfurt, Germany.

Should I upgrade to n8n 3.0 now?

Only on a test instance for now. 3.0 is an early-access beta with stable due in early December 2026, and the upgrade cannot be rolled back. Self-hosters on npm must move to Docker first.

Key takeaways

• n8n bills per workflow execution, not per step, which makes it the cheapest of the big three for long, multi-step workflows.
• Cloud prices are euro-only: Starter 20€ and Pro 50€ a month billed annually. Business (667€) is self-hosted only.
• The self-hosted Community Edition is free with no execution caps, but it is fair-code and lacks team and governance features.
• Check the limits beyond executions: Starter caps runs at 5 minutes and keeps logs for 7 days.
• Fees are non-refundable, and a lapsed trial instance is deleted. Export your workflows within 90 days.
• n8n 3.0 is in beta until early December 2026. It is a one-way upgrade and requires Docker for self-hosting.