n8n

What it is

n8n is a workflow automation platform for technical teams: a visual canvas where each step is a node, with JavaScript and Python code steps when the canvas runs out, AI agent nodes that connect any LLM to tools and memory, and more than 1,000 integrations. You run it on n8n’s cloud or on your own infrastructure. The public template library lists more than 12,000 community workflows. It sits between Zapier, which is easier for non-technical users, and enterprise iPaaS like Workato, which costs far more.

It is source-available, not open source. The self-hosted Community Edition is free, but it ships under n8n’s Sustainable Use License, which limits use to “your own internal business purposes” or non-commercial use. Files marked .ee need a paid license. Our previous version of this page called the Community Edition MIT-licensed. That was wrong, and it matters to anyone planning to build a product on n8n.

The company is independent and well funded. It raised a $180M Series C led by Accel on October 9, 2025, at a $2.5B valuation, with NVIDIA’s NVentures participating, and reported 10x revenue growth in the year before. In May 2026 SAP bought shares in a secondary sale that valued n8n at $5.2B, and signed a multi-year agreement to embed n8n in SAP’s Joule Studio agent builder.

Why it shows up in RevOps, Legal Ops and Recruiting stacks

• Billing counts runs, not steps. An execution is one run of a whole workflow however many steps it has. A 20-step enrichment-and-routing flow that fires 500 times a month uses 500 executions on n8n and 10,000 tasks on a per-step meter. That one design choice is why teams move here when their workflows get longer.
• MCP works in both directions. An instance-level MCP server lets Claude Desktop, Claude Code, ChatGPT, Cursor and other clients search workflows, run the ones you have exposed, and, from n8n 2.13, build and edit workflows and data tables. The MCP Server Trigger node turns a single workflow into a purpose-built MCP server, and the MCP Client Tool node lets an n8n agent call outside MCP servers.
• Data can stay in your perimeter. Self-hosted, customer records never leave your infrastructure. On n8n’s cloud, data sits in Frankfurt, which settles the question for EU-based Legal Ops and Recruiting teams working with candidate and contract data.

Pricing reality

n8n.io/pricing quotes in euros only, checked 2026-09-11. Every plan includes unlimited users, unlimited workflows and every integration. Prices below are per month, billed annually; monthly billing costs 20% more (the page frames annual as “Save 17%”).

• Community Edition — free, self-hosted. Unlimited executions, but no SSO, Git version control or environments.
• Starter — €20/month (about $23). 2,500 executions, 5 concurrent, 1 shared project, 2,300 AI Assistant credits. Cloud only.
• Pro — €50/month. 10,000 executions, 20 concurrent, 3 shared projects, admin roles, global variables, workflow history, execution search, up to 13,700 AI credits. Cloud only.
• Business — €667/month. 40,000 executions, 6 shared projects, SSO/SAML/LDAP, Git version control, dev/prod environments, 30 days of insights. Self-hosted only, aimed at companies under 100 employees, forum support only. Overage is €4,000 per extra 300,000 executions.
• Enterprise — custom. Cloud or self-hosted, 200+ concurrent executions, log streaming, external secret stores, 365 days of insights, support with an SLA.
• Start-up Plan. Companies under 20 employees can get 50% off Business, which works out to about €333/month.

The plans changed this year. The previous version of this page listed Starter at $24 with 5,000 executions and Pro at $50. Starter now includes 2,500 executions, and Business at €667 is a new tier.

What teams actually pay: Vendr’s data across 51 purchases shows a median contract of $700 a year, which is Pro billed annually. The expensive jump is from Pro to Business. Going from €600 to €8,004 a year buys SSO and Git-backed environments, and it also means moving from n8n’s cloud to hosting it yourself. Self-hosting costs about $40–80 a month in infrastructure on a mainstream cloud (our estimate: one small VM, managed Postgres, Redis for queue mode), plus engineer time.

Best for

• Best for: RevOps and GTM engineers at 20–200 person companies who build multi-step automation that touches CRM, enrichment and messaging, and who want code steps and agent nodes in the same tool. ROI is clearest once workflows run past 10 steps, because the per-execution meter keeps them cheap.
• Legal Ops and Recruiting teams that need contract or candidate data self-hosted or kept in the EU.
• Teams standardising on Claude or Cursor that want one orchestration layer those assistants can call through MCP.
• Not for marketing or sales teams with no one who can read JSON or debug a failed HTTP node. Nor for anyone planning to resell n8n-powered automation as their own product, which the license forbids without a separate agreement.

Versus the alternatives

• Zapier — the market-share leader, with the widest connector catalogue and the gentlest learning curve, from $19.99/month. Pick it when the builders are non-technical and flows are short. Pick n8n once flows get long, because Zapier bills every step as a task.
• Microsoft Power Automate — the other share leader, bundled into many Microsoft 365 agreements. Pick it when the company runs on Microsoft 365 and IT wants automation inside that tenant’s governance.
• Make — the cheapest paid entry, Core at $9/month, now billed in credits, with a strong visual builder for complex branching. Pick it when budget is tight and nobody needs to self-host.
• Gumloop — the fastest-growing AI-native entrant, with a $50M Series B led by Benchmark in March 2026. Pro is $37/month for 20,000 credits. Pick it for teams with no engineers running document- and model-heavy agent pipelines. Pick n8n when you want code steps, self-hosting and a pricing unit that doesn’t meter model calls.
• Activepieces — the open-source alternative: MIT core, per-run billing like n8n, Plus at $16/month. Pick it when the license has to be genuinely open source, or when you intend to embed automation in a product you sell.
• Workato — enterprise iPaaS with a Vendr median of $64,544 a year. Pick it for governed CRM-to-ERP order-to-cash at enterprise scale, where n8n’s forum-support Business tier is too thin.

Watch-outs

• The license is not open source. The Sustainable Use License allows internal business use and non-commercial distribution only. Guard: if n8n will run workflows for your customers, or ship inside a product you sell, get n8n’s Embed or commercial terms in writing before you build. If that is the plan from day one, evaluate Activepieces as well.
• The Pro-to-Business step is a price jump and a platform move. SSO, Git version control and environments are Business features, and Business is self-hosted only. Guard: at the pilot stage, decide whether security will require SSO within 12 months. If so, budget for Business, or the Start-up Plan if you qualify, from the start rather than building on Cloud Pro and migrating later.
• Polling triggers use up executions. n8n’s own FAQ notes that a schedule running every 5 minutes uses about 8,600–8,900 executions a month. That is more than Starter’s 2,500 from a single workflow. Guard: use webhooks wherever the source app supports them, and check execution counts in the usage page during the first two weeks.
• Instance-level MCP exposure is not scoped per client. Every connected client, Claude Desktop and ChatGPT alike, sees every workflow you have enabled for MCP. Guard: enable only workflows that are safe for any connected assistant to trigger, and keep write-heavy or destructive workflows off MCP or in a separate instance.
• Self-hosted paid plans contact n8n daily. Business and Enterprise license keys ping n8n’s license server once a day with production execution counts, and telemetry is on by default. Guard: allow outbound traffic to the license server in your network policy, turn telemetry off through the documented environment variables, and tell your security review both facts before they find them.