n8n Pricing 2026: Free Self-Hosted vs Cloud from €20

n8n Community Edition is free if you self-host it. Cloud starts at €20/month (Starter, annual rate) for 2,500 executions, then Pro at €50 and Business at €667. After two years of production self-hosting, the real cost is never $0 — you still pay for the server and the person who keeps it up. Compare that against Zapier or Make in the AI Cost Calculator.

Why n8n Looks So Cheap: It Meters Runs, Not Steps

Before any price comparison, understand what each platform counts. This single difference explains most of the gap, and almost no pricing guide spells it out.

Take one real workflow: a nightly job that pulls 50 new rows from a database, enriches each through an API, and writes them to a CRM. In n8n that is 1 execution. In Zapier it is roughly 400 tasks. In Make, roughly 400 credits. Run it daily and n8n charges you 30 executions a month against your 10,000, while Zapier charges 12,000 tasks — which is above their 2,000-task tier and into $733.50/month territory.

This is why "n8n Pro is €50 for 10,000 and Zapier Professional is $73.50 for 2,000" understates the difference rather than overstating it. Compare units before you compare prices.

n8n Plans Overview

Every figure below was checked against n8n.io/pricing and n8n's support documentation on 5 August 2026. n8n publishes in euros and shows the annual-billed rate by default; the toggle on their page advertises 17% off for annual, so monthly billing runs higher. n8n does not publish the monthly figure, so we do not quote one.

Community Edition: Free (Self-Hosted)

• 100% free and open-source (Fair-code license)
• Unlimited workflows and executions
• All 400+ integrations included
• No execution-based billing
• You host it on your own infrastructure
• Community support only (forums, Discord)
• No built-in SSO, audit logs, or enterprise features

Honest take: This is genuinely the best free automation tool available. Period. You get everything -- every node, every integration, unlimited executions. The catch is you need to run it somewhere, keep it updated, handle backups, and fix it when it breaks at 2 AM. And it will break at 2 AM eventually.

Cloud Starter: €20/month (billed annually)

• 2,500 workflow executions per month
• Unlimited workflows, steps and users
• 5 concurrent executions
• 1 shared project
• 2,300 AI Assistant credits per month
• All integrations included
• n8n-hosted (no infrastructure management)
• Execution history (7 days)

Honest take: 2,500 executions sounds like a lot until you realize a single workflow that runs every 5 minutes burns through 8,640 executions per month. Even a workflow that runs hourly uses 720. Starter works for a handful of low-frequency workflows. It does NOT work for any real automation setup. Note what changed: you can now switch on as many workflows as you like on Starter — the cap that bites is the frequency they fire at, not their number.

Cloud Pro: €50/month (billed annually)

• 10,000 workflow executions per month (Pro-1)
• Unlimited workflows, steps and users
• 20 concurrent executions
• 3 shared projects
• 5,700 AI Assistant credits per month
• Execution history (14 days)
• Debug executions, global variables, workflow history, execution search
• Admin roles
• A Pro-2 step exists at 50,000 executions / 50 concurrent / 13,700 AI credits — n8n does not publish its price on the pricing page; you size it in the admin panel

Honest take: This is the minimum viable plan for most businesses. 10,000 executions handles a genuinely useful set of workflows now that step count is free. The trap is not overage billing — it is error executions, which consume allowance exactly like successful ones. We had a client burn through most of a month's allowance in two days because a webhook fired in a retry loop and each failure counted.

On running past your allowance: n8n's documentation says additional executions are purchased — self-serve in the admin panel for monthly subscribers, or by contacting n8n for annual ones. It is not an automatic per-execution surcharge. Older guides (including an earlier version of this one) quote "$0.007 per additional execution"; we could not find that figure anywhere on n8n's site or docs in August 2026, so we no longer repeat it.

Cloud Business: €667/month (billed annually)

• 40,000 workflow executions per month
• Self-hosted deployment only
• 30 concurrent executions
• 6 shared projects
• SSO / SAML / LDAP
• Git-based version control and separate environments
• 30 days of insights
• Overage: n8n quotes €4,000 for an additional bucket of 300,000 executions (≈€0.013 each)

Honest take: Business is the tier most write-ups miss, and the jump from €50 to €667 is steep. You are not buying executions at that price — 40,000 executions is only 4x Pro-1 for 13x the money. You are buying SSO, Git version control, and environments. If your blocker is an IT department that requires SAML, this is the cheapest door. If you just need volume, self-hosting Community gets you unlimited executions for the price of a VPS.

Cloud Enterprise: Contact sales

• Custom execution volume
• Unlimited shared projects
• 200+ concurrent executions
• Hosted by n8n or self-hosted
• External secret store integration and log streaming
• 365 days of insights
• Dedicated support with an SLA

Honest take: n8n moved Enterprise off workflow-based pricing onto execution-based pricing in August 2025, so any quote you were given before then does not describe the current model. There is no published starting price and we will not invent one — the number we used to print here ($300/month) is from the old model and we have removed it. If you need external secret stores or log streaming, you are in this tier regardless of volume.

Self-Hosted vs Cloud: The Real Math

When self-hosted wins

• You run 10,000+ executions per month (Cloud Pro starts losing to self-hosted here)
• You have someone technical who can manage a VPS and Docker
• You need full control over data residency and retention
• You want to run n8n with local AI models (LLaMA, Mistral) without API costs

When Cloud wins

• You have no DevOps capability (no one to SSH into a server)
• You value reliability over cost savings (Enterprise is the tier that carries a contractual SLA)
• You run fewer than 5,000 executions per month
• You need quick setup (Cloud is running in 2 minutes, self-hosted takes 1-2 hours minimum)

n8n vs Zapier vs Make: Price Comparison

This is where n8n's value proposition really shines:

All figures verified 5 August 2026. Remember the units differ — see the table at the top of this page before reading across rows.

Reading this correctly: Make Core at $12 for 10,000 credits looks like it beats n8n Pro at €50 for 10,000 executions. It does not, because a credit is a step and an execution is a whole run. For a five-step workflow, Make's 10,000 credits buy you 2,000 runs; n8n's 10,000 executions buy you 10,000. For a fifty-step workflow the gap is tenfold again. Make is genuinely cheaper only for workflows that are one or two steps long.

Zapier's overage is worth knowing before you enable it: with pay-per-task on, you pay 1.25x your base rate on annual plans and 2.5x on monthly. Going over on a monthly Zapier plan is the most expensive automation you can buy.

For the full feature comparison, read our n8n vs Zapier comparison or n8n vs Make comparison.

n8n AI Features & AI Agent Pricing (2026)

n8n launched its AI Agent node in 2024 and has since become one of the cheapest ways to run tool-calling LLM agents in production. But there are two separate AI meters here, and conflating them is the most common mistake in every other guide on this query.

So running AI costs you nothing extra from n8n; being helped to build it does. If you are pricing a production deployment, only the first row matters — the credits are a development-time convenience, and they reset monthly whether or not you used them.

What AI Features n8n Includes (All Plans)

• AI Agent node: Multi-step tool-calling agent with memory, OpenAI function-calling format
• LLM provider nodes: OpenAI, Anthropic, Azure OpenAI, Google Gemini, Groq, Mistral, Cohere, DeepSeek
• Local LLM nodes: Ollama (LLaMA 3.3, Mistral, Qwen, DeepSeek-R1 locally hosted)
• Vector stores: Pinecone, Supabase, Qdrant, Weaviate, in-memory
• Embeddings: OpenAI, Cohere, HuggingFace, local
• Document loaders: PDF, CSV, JSON, web scraping, databases
• Output parsers: Structured JSON, auto-fixing, item lists

Real LLM Token Costs (Per Workflow Run)

• Simple classification (gpt-4o-mini, ~500 tokens): $0.0001
• Email draft (gpt-4o-mini, ~2,000 tokens): $0.001
• RAG Q&A with retrieval (gpt-4o, ~4,000 tokens): $0.04
• 3-tool agent with memory (claude-sonnet-4, ~8,000 tokens): $0.12
• Complex multi-step research agent (claude-opus-4, ~20,000 tokens): $0.60
• Same workflow on Ollama + LLaMA 3.3 70B self-hosted: $0 (compute only)

Monthly AI Cost Examples (From Our Clients)

• Lead-qualification workflow (500 leads/mo × $0.002 avg): $1/mo in tokens
• Customer support triage (3,000 tickets × $0.01): $30/mo in tokens
• Content-generation pipeline (100 articles × $0.40): $40/mo in tokens
• Research agent (200 runs × $0.50): $100/mo in tokens
• Same workloads on self-hosted Ollama GPU ($80/mo VPS): $80/mo flat, unlimited runs

n8n AI Agent vs Competitors

• n8n AI Agent: $0 platform fee + API tokens. Visual builder, 400+ integrations.
• Zapier AI Actions: Uses your Zapier task quota ($29.99+/mo). Locked to Zapier's LLM selection.
• Make AI Tools: Uses Make operations quota ($10.59+/mo). Decent model coverage.
• LangChain (custom code): $0 framework + dev time (8-12 hours per agent) + infrastructure. Most flexible but expensive in labor.
• Langflow / Flowise: Open-source agent builders. Similar to n8n but narrower integration catalog.

Honest take: For 80% of agent use cases (tool-calling, RAG, multi-step workflows with APIs), n8n is the cheapest option considering total cost of ownership. We switched 6 of our internal LangChain agents to n8n in Q1 2026 and cut our dev time on new agents from ~10 hours to ~1 hour each. For truly novel agent architectures (planner-executor patterns, custom reasoning chains), LangChain code still wins.

Our Setup and What We Pay

We run a hybrid setup: self-hosted n8n for internal workflows and heavy-volume client automations, plus n8n Cloud for clients who do not want to manage infrastructure.

• Internal: Self-hosted on Hetzner CX32 ($12/month) + managed PostgreSQL ($15/month). Running 40+ workflows, ~200,000 executions/month. Total: $27/month.
• Client deployments: Mix of self-hosted (for tech-savvy clients) and Cloud Pro/Enterprise (for others). Average client pays $50-300/month depending on volume.

Our internal instance runs about 200,000 executions a month. There is no Cloud tier that covers that without a sales conversation — Pro-2 tops out at 50,000 and Business at 40,000. The closest published number to price it against is Business overage, where n8n quotes €4,000 for an extra 300,000 executions. At that rate our internal volume alone is roughly €2,700/month of execution allowance, against $27/month of hardware. That gap is the entire argument for self-hosting, and it only exists because we already had someone who could run it.

Reverse the point honestly: at 2,000 executions a month, self-hosting costs $130-230 all-in and Starter costs €20. Below roughly 10,000 executions, Cloud wins outright and self-hosting is a hobby.

Hidden Costs and Gotchas

• Self-hosted updates: n8n releases updates every 1-2 weeks. If you fall behind, you risk security issues and missed bug fixes. Budget 30-60 minutes per update (they usually go smoothly, but test first).
• Cloud execution counting: Error executions still count toward your limit. A broken webhook that fires and errors 1,000 times burns 1,000 executions. Set up error-handling sub-workflows.
• External API costs: n8n itself is just the orchestrator. If your workflows call OpenAI, Twilio, or other paid APIs, those costs are separate and can add up fast.
• Worker mode for self-hosted: High-volume self-hosted deployments need worker mode (separate process for execution). This roughly doubles your server requirements.

Frequently Asked Questions

Is n8n really free?

The software is free under n8n's fair-code Sustainable Use License. You can download it and run it on your own computer or server at no licence cost. Cloud hosting starts at €20/month billed annually. The honest answer is that n8n is free the way a puppy is free.

What is the best n8n plan for a small business?

If you have someone technical who will actually own the server: self-hosted on a $10-15/month VPS. If you do not: Cloud Pro at €50/month. Starter is worth a look now that workflow counts are unlimited — but 2,500 executions still runs out fast if anything fires on a schedule tighter than hourly.

Can I switch from self-hosted to Cloud (or vice versa)?

Yes. n8n lets you export and import workflows. The migration is straightforward -- workflows transfer cleanly. Credentials need to be re-entered manually for security reasons. We typically complete a migration in 1-2 hours.

Does n8n have a free trial for Cloud?

Yes — 14 days with access to Pro features, no credit card required. The trial includes 2,500 executions and 800 AI Assistant credits. That is usually enough to build 3-5 workflows and find out whether your real execution volume fits inside Starter or forces you to Pro.

How current is this page?

Every n8n figure here was checked against n8n.io/pricing and n8n's support documentation on 5 August 2026, and the Zapier and Make figures against their own pricing pages on the same day. n8n has changed its pricing model twice since 2024 — most guides ranking for this query still describe active-workflow limits that were removed a year ago. If you are reading this well after that date, check the source links before you budget against it.