Last updated: 6 October 2026
Quick Answer: n8n Cloud starts at 20 euros a month for Starter with 2,500 workflow executions, billed annually. n8n lists its prices in euros. Pro is 50 euros for 10,000 executions, Business is 667 euros and self-hosted only, and the Community Edition is free to self-host.
n8n bills per workflow run, not per step, so workflow design matters more than step count.
Most reads of n8n pricing stop at the plan table. Once a model is involved, tokens and build time dwarf it.
How much does n8n cost in 2026?
Four paid tiers and one free one. Starter and Pro are hosted by n8n; Business and Enterprise are licences you run yourself. Every tier on n8n's pricing page includes unlimited users, workflows and integrations, so you buy executions, not seats.
| Plan | Price a month, annual | Executions | Runs on | Concurrent | Extras |
|---|---|---|---|---|---|
| Starter | 20 euros | 2,500 | n8n Cloud | 5 | 1 project, Assistant credits |
| Pro | 50 euros | 10,000 | n8n Cloud | Up to 50 | 3 projects, admin roles |
| Business | 667 euros | 40,000 | Self-hosted only | Unlimited | SSO, Git, environments |
| Enterprise | Contact sales | Custom | Cloud or self-hosted | 200+ on Cloud; unlimited self-hosted | Log streaming, SLA support |
| Community Edition | Free | No quota stated | Self-hosted | Unlimited | Sustainable Use License |
n8n pricing plans as listed on n8n.io on 6 October 2026, in euros and at the annual rates. Self-hosted concurrency is listed unlimited, bounded by your hardware and queue mode.
What does each n8n Cloud plan include?
n8n Cloud pricing covers two plans. Starter gives 2,500 executions, one project, five concurrent runs, a five-minute ceiling and seven days of logs. Pro raises that to 10,000 executions, three projects, 20 concurrent runs rising to 50 on the 50,000 bucket, a 40-minute ceiling and 30 days of logs.
What counts as an n8n execution?
n8n's definition is blunt: "An execution is a single run of your entire workflow. It doesn't matter how many steps are in the workflow or how much data it processes. It's still a single execution." Branches, filters and loops inside one workflow add nothing.
Frequency is what costs you. A workflow on a 15-minute schedule fires roughly 2,880 times a month, clearing Starter's allowance on its own. n8n counts each agent turn as an execution, so a busy chat agent hits the ceiling first. Manual test runs, sub-workflows and error workflows are excluded from the concurrency limit.
Is self-hosted n8n really free?
Partly. There is no separate n8n on-premise pricing tier: self-hosting is either the free Community Edition or a paid Business or Enterprise licence on your own servers. n8n self-hosted pricing charges for the licence, never the hosting, which stays your line item.
What can you do with the n8n Community Edition?
n8n Community Edition pricing is zero, and the limits are both legal and technical: n8n's matrix excludes Community from SSO and LDAP, Git, environments and shared projects. The Sustainable Use License FAQ says n8n is free where workflows are only created or modified inside your own organization. It permits:
- running your business on it, plus personal and research use
- building client automations clients cannot edit themselves
- charging for workflow creation, setup, maintenance and consultancy
- using n8n inside your product, where users trigger flows you built
It bars hosting n8n as a service where clients build workflows, letting external users configure workflows through your product by any route including a custom UI, the API, MCP or an agent, forking the code into a rival product, or white-labelling it.
What does self-hosting n8n cost in infrastructure and upkeep?
n8n does not price this; these are your cloud and payroll lines.
- A cluster sized for peak concurrency, not average.
- A Postgres instance, managed or self-run.
- Redis and queue mode once one worker stops keeping up.
- Tested backups, upgrades, and on-call time when a workflow fails silently.
Price those before comparing self-hosting with Cloud.
What do n8n Business and Enterprise add, and who needs them?
Business is where self-hosting becomes governable: SSO and LDAP, Git version control, environments, six shared projects and 40,000 executions. It is self-hosted only; n8n lists Cloud availability as not yet. n8n Enterprise pricing is quote-only and adds 200 or more concurrent executions on Cloud, log streaming and support with an SLA.
Business is self-serve, and n8n states dedicated support is Enterprise only. Its pricing FAQ puts Business overage at 4,000 euros per extra bucket of 300,000 executions, invoiced 45 days later. Companies under 20 employees can apply for a Start-up Plan at 50% off Business.
How does n8n AI pricing work for the Assistant?
n8n AI pricing arrives as Assistant credits, not a separate bill. Starter includes 1,600 a month and Pro up to 9,600. The balance refreshes monthly, unused credits do not roll over, and burn depends on what you build.
Is n8n cheaper than Make or Zapier?
Usually, once a workflow has more than two or three actions, because the three meter differently. n8n counts one execution per run. Zapier counts a task per successful action, excluding triggers, polling, filters and formatters. Make counts a credit per module run, one credit per operation on non-AI apps.
| Platform | Billing unit | A trigger plus four actions | 2,000 runs a month |
|---|---|---|---|
| n8n | One execution per run | 1 execution | 2,000 executions |
| Zapier | One task per action, triggers excluded | 4 tasks | 8,000 tasks |
| Make | One credit per module, trigger included | 5 credits | 10,000 credits |
Counting rules as documented by each vendor on 6 October 2026.
Make's operations guide counts the trigger as one operation, and Zapier notes many actions cost more than one task, so 8,000 is a floor. Where n8n loses is connector breadth and upgrade ownership, a trade our n8n alternatives guide works through.
What mistakes push your n8n bill higher than it needs to be?
- Polling on a schedule where a webhook would do: every poll is a full execution.
- Ignoring the 17% annual discount on a plan you mean to keep.
- Buying Business for SSO when Pro's execution ceiling was the constraint.
- Calling the Community Edition free before pricing Postgres and upkeep.
What does an AI workflow built on n8n really cost?
Three lines, and the plan is usually the smallest. The model bill scales with tokens per run, build time with how many systems it touches.
On OpenAI's API pricing page, read on 6 October 2026, gpt-6.1-sol costs $2.00 per million input tokens and $10.00 per million output tokens, while gpt-6-luna costs $0.10 and $0.50.
Take a support-triage workflow running 2,000 times a month at 2,000 input and 500 output tokens a run: 4 million input tokens and 1 million output.
| Line | Volume | gpt-6.1-sol | gpt-6-luna |
|---|---|---|---|
| Input tokens | 4,000,000 | 8.00 | 0.40 |
| Output tokens | 1,000,000 | 10.00 | 0.50 |
| Model total | 18.00 | 0.90 |
Arithmetic from the rates above, in US dollars.
At that volume the model is a rounding error and the plan dominates. Push it to 200,000 runs on the larger model and the token line alone is roughly a hundred times that, overtaking any n8n plan. Our AI workflow cost breakdown itemises discovery, integrations, evaluation and handover.
When does a custom AI workflow make more sense than an n8n plan?
When the limit you hit is not the execution count. Four signals say the platform is the wrong shape rather than the wrong size: agents needing authority to act on production systems, data that must stay inside your network, retry and approval logic that outgrew the canvas, and the licence bar on your own users building workflows.
That last one decides more migrations than price does. If you are shortlisting firms that build custom AI workflows for enterprises, judge them on integrations already shipped, how they evaluate agent output against a test set, and whether they hand over a system you can run. Compare partners with our AI automation consultants guide.
What should an AI workflow development quote include?
- Scope as the decisions the workflow owns.
- Every integration named, with read or write access.
- Model choice and a token budget per run.
- An evaluation plan, and the fallback when the model is wrong.
How Origins AI builds AI workflows when n8n is not enough
Origins AI (originshq.com) is a US-based AI engineering company that builds custom AI workflows, agents and LLM integrations for product teams. Its AI services page covers AI product and model development, consulting, automation solutions, generative AI, solution architecting and OpenAI integration work, and describes encryption at rest and in transit, secure authentication and least-privilege access.
Where a canvas runs out, the agentic automation page describes agents given authority to execute rather than only recommend, across approvals, triage and routing, monitoring and data operations, with escalation thresholds set during agent design.
Origins AI reports 1,200 hours saved annually on average and a decision cycle falling from 45 minutes to 3 minutes, and does not publish a rate card. Its services page lists four ways to engage: an embedded AI team, a scoped project, time and materials, or build-operate-transfer. If an n8n plan covers your workflows, stay on it.
Talk to an engineer
Send the workflow you want priced and we will itemise it: the plan, the tokens per run, the build. Book a call with an engineer.