n8n looks free — the code is on GitHub and you can run it on a $12 server. It also has cloud plans from €20 a month, a €667 self-hosted Business licence, and an AI model bill that none of those prices include. So "how much does n8n cost?" has three answers, and which one is yours depends on how many times your workflows run, not on how many steps they have.
This guide is for founders, operators and small teams deciding where to run AI automations: n8n Cloud, self-hosted n8n, Zapier or Make. We price one realistic AI workflow at 1,000, 10,000 and 100,000 runs a month on each, using the prices published on each vendor's own page on October 9, 2026. It does not cover enterprise contracts (those are quotes, not prices), agencies reselling n8n to clients (a licensing question, covered briefly below), or n8n's own AI Assistant credits, which help you build workflows rather than run them.
The answerOne AI workflow, priced four ways
Our test workflow is ordinary: a trigger (a new form submission or email), seven nodes after it — look up the contact, call an AI model to classify and draft a reply, write to a sheet, post to Slack, and so on — and one model call of 2,000 input tokens and 300 output tokens per run.
n8n Cloud
- 1,000 runs
- €24 Starter ($27) [€20]
- 10,000 runs
- €60 Pro ($67) [€50]
- 100,000 runs
- Beyond Pro — quote
n8n self-hosted (Community)
- 1,000 runs
- $12 server
- 10,000 runs
- $12 server
- 100,000 runs
- $24 server
Zapier Professional
- 1,000 runs
- 7,000 tasks → $193.50 [$129]
- 10,000 runs
- 70,000 tasks → $733.50 [$489]
- 100,000 runs
- 700,000 tasks → $2,999 [$1,999]
Make Core
- 1,000 runs
- 8,000 credits → $9 (10k-credit tier)
- 10,000 runs
- 80,000 credits
- 100,000 runs
- 800,000 credits
| Option | 1,000 runs | 10,000 runs | 100,000 runs |
|---|---|---|---|
| n8n Cloud | €24 Starter ($27) [€20] | €60 Pro ($67) [€50] | Beyond Pro — quote |
| n8n self-hosted (Community) | $12 server | $12 server | $24 server |
| Zapier Professional | 7,000 tasks → $193.50 [$129] | 70,000 tasks → $733.50 [$489] | 700,000 tasks → $2,999 [$1,999] |
| Make Core | 8,000 credits → $9 (10k-credit tier) | 80,000 credits | 800,000 credits |
Source: CalculatorAI · calculatorai.app · n8n.io/pricing · zapier.com/pricing · make.com/en/pricing · digitalocean.com/pricing/droplets · ECB rate 1.1186 (Oct 8, 2026)
Three things to read off it before the details. n8n Cloud's price moves with runs, Zapier's with runs × steps, Make's with runs × modules. The self-hosted row barely moves at all, because nobody is counting. And the 100,000-run column is where n8n Cloud's published plans stop: Starter includes 2,500 executions and Pro 10,000, and above that n8n's page points to its Business licence or an Enterprise quote.
How n8n countsOne run is one execution, however long
n8n defines an execution as "a single run of your entire workflow", regardless of how many steps it has or how much data it moves. A three-node workflow and a fifty-node workflow cost the same per run on n8n Cloud.
That one rule is most of n8n's pricing story, because the other two count differently:
- Zapier counts a task "whenever Zapier successfully completes a unit of work" — every action step that succeeds. Triggers, filters, paths and the Formatter are free; failed steps are free.
- Make counts a credit for every module that runs, "like adding a Google Sheet row or fetching Gmail account"; routers and error handlers are free.
Going from 3 to 20 steps quadruples the Zapier bill at the same 10,000 runs; the n8n Pro fee does not move.
Show these figures as a table
| Value ($ a month) | |
|---|---|
| 3 steps on Zapier — 20,000 tasks tier | 189 |
| 8 steps on Zapier — 100,000 tasks tier | 489 |
| 20 steps on Zapier — 200,000 tasks tier | 769 |
| Any step count on n8n Pro — €50, 10,000 executions | 56 |
Source: CalculatorAI · calculatorai.app · zapier.com/pricing · n8n.io/pricing
So the longer and busier the workflow, the more n8n's per-run counting is worth. For a two-step "new form → email me" automation that runs 300 times a month, the difference is a few dollars and Zapier's free and low tiers are perfectly reasonable.
The model billWhat the AI step adds on every platform
None of these plans pays for the AI model inside your workflow. You connect your own OpenAI, Anthropic or Google key and pay the provider per token, wherever the workflow runs. n8n Cloud's plans include "AI credits", but those are for n8n's Assistant — the helper that builds workflows for you — not for the model calls your workflow makes.
GPT-6 Luna ($0.10 / $0.50)
- Per run
- $0.00035
- 1,000 runs
- $0.35
- 10,000 runs
- $3.50
- 100,000 runs
- $35
Gemini 3.8 Flash ($0.75 / $3.75)
- Per run
- $0.0026
- 1,000 runs
- $2.63
- 10,000 runs
- $26
- 100,000 runs
- $263
Claude Haiku 4.5 ($1 / $5)
- Per run
- $0.0035
- 1,000 runs
- $3.50
- 10,000 runs
- $35
- 100,000 runs
- $350
Claude Sonnet 5.5 ($2 / $10)
- Per run
- $0.0070
- 1,000 runs
- $7
- 10,000 runs
- $70
- 100,000 runs
- $700
| Model | Per run | 1,000 runs | 10,000 runs | 100,000 runs |
|---|---|---|---|---|
| GPT-6 Luna ($0.10 / $0.50) | $0.00035 | $0.35 | $3.50 | $35 |
| Gemini 3.8 Flash ($0.75 / $3.75) | $0.0026 | $2.63 | $26 | $263 |
| Claude Haiku 4.5 ($1 / $5) | $0.0035 | $3.50 | $35 | $350 |
| Claude Sonnet 5.5 ($2 / $10) | $0.0070 | $7 | $70 | $700 |
Source: CalculatorAI · calculatorai.app · CalculatorAI model catalogue, verified against developers.openai.com, ai.google.dev and platform.claude.com
At 100,000 runs on Claude Sonnet 5.5, the model costs $700 a month — more than the $24 server the workflow runs on, nearly thirty times over. That is the general shape: at low volume the platform is the bill; at high volume the model is. The levers there are model choice and token count, not the automation tool. Our guides to what AI actually costs and prompt caching cover both, and a workflow that loops — an AI agent node calling tools several times per run — is priced step by step in how much an AI agent costs per month.
Worked example10,000 runs a month, all in
An AI lead-triage workflow, 10,000 runs a month on Gemini 3.8 Flash
Eight steps per run, one model call of 2,000 tokens in and 300 out. Monthly billing everywhere, euro prices converted at the ECB rate of 1.1186.
The model cost is identical in all three columns, which is the point: the choice of platform moves this bill by up to 20×, the choice of model by another 20× (GPT-6 Luna would make it $3.50, Claude Sonnet 5.5 $70).
Self-hostingFree software, not a free service
The self-hosted Community edition has no execution limit; n8n's documentation lists none, and it is the same product as the cloud, minus some team features. What you pay instead:
- A server. n8n runs comfortably for small teams on a 2 GB machine; we used DigitalOcean's 1 vCPU / 2 GB droplet at $12 a month for up to 10,000 runs and the 2 vCPU / 4 GB one at $24 for 100,000. Heavier workflows (large files, many parallel runs) need more.
- Your time. Updates, backups, a database that eventually outgrows SQLite, monitoring, the night a webhook stops answering. At $50 an hour, self-hosting is cheaper than n8n Cloud Pro billed monthly only while upkeep stays under about 1.1 hours a month — it rarely does in the first month, and often does after that.
- The features you do not get. The Community edition leaves out, among others, workflow and credential sharing, projects, environments, Git version control, external secrets, log streaming and SSO. A solo builder rarely misses them; a team of five usually does, which is what the self-hosted Business licence (€667 a month, billed yearly, 40,000 executions included) is for.
Nobody wants to run a server
Under 10,000 runs a month, a small team, and no one whose job includes updates and backups. €24–€60 a month buys that away.
Volume is high or data must stay in-house
Tens of thousands of runs, data that cannot leave your infrastructure, and someone comfortable with Docker and Postgres.
The other sideWhen Zapier or Make is the better buy
Per-run pricing is not automatically cheaper. Zapier's Free plan includes 100 tasks a month and its Professional plan starts at $19.99 a month billed yearly for 750 tasks; Make's Free plan includes 1,000 credits and Core starts at $9. For a handful of short automations, both cost less than n8n Cloud Starter, and both come with thousands of ready-made app connections that a non-technical teammate can set up without touching JSON.
Two more things an n8n enthusiast tends to skip:
- Failed steps. Zapier's pricing page says failed actions are not counted as tasks. Retries are runs too, so a flaky workflow that retries itself also raises its own execution or credit count — measure your failure rate before you compare.
- Polling. A Make scenario that checks for new data on a schedule uses a credit every time its trigger runs, even when there is nothing new — a 5-minute schedule is about 8,600 checks a month before any real work. Zapier's triggers and polling are free.
So the honest rule: few runs, short workflows, non-technical owners → Zapier or Make. Many runs, long workflows, someone technical → n8n.
MethodWhere these numbers come from
Prices were read on October 9, 2026 from n8n.io/pricing (Starter €20 / Pro €50 / Business €667 a month billed yearly; 2,500 / 10,000 / 40,000 executions — the page shows yearly prices, and the monthly-billing figures of €24 and €60 follow n8n's 17% yearly discount), zapier.com/pricing (Professional tier table, yearly and monthly), make.com/en/pricing (entry prices and credit rules) and digitalocean.com/pricing/droplets ($12 and $24 Basic droplets). Euro prices are converted at the ECB reference rate of 1.1186 USD per EUR (October 8, 2026); taxes are excluded. Model prices come from our calculator catalogue, verified against the providers' pricing pages on October 1, 2026.
The workflow is an assumption chosen to be typical: 8 steps, so 7 Zapier tasks (the trigger is free) and 8 Make credits per run, one model call of 2,000 input and 300 output tokens, no caching, no retries. Where this biases the result: retries would add runs on every platform (Zapier does not count failed actions); prompt caching or a batch API would lower the model rows; the self-hosted rows exclude labour, backups and a managed database, so they are floors, not totals. The calculation script is reproducible from these inputs.
FAQn8n pricing
Is n8n free?
The self-hosted Community edition is free and has no execution limit, under the Sustainable Use License: free for your own internal business, personal or non-commercial use. You still pay for the server, your time and any AI model calls. n8n Cloud, where n8n runs it for you, starts at €20 a month billed yearly (€24 monthly) for 2,500 executions.
How does n8n count executions?
One execution is one complete run of a workflow, whatever its number of steps or the amount of data processed. That is why long workflows are much cheaper on n8n than on Zapier, which counts each successful action as a task.
Do n8n's AI credits pay for OpenAI or Claude calls in my workflows?
No. The AI credits on n8n Cloud plans are for n8n's Assistant, which helps you build workflows. Model calls inside a workflow use your own API key and are billed by OpenAI, Anthropic, Google or whichever provider you connect.
Is n8n cheaper than Zapier?
For busy, multi-step workflows, by a wide margin: our 8-step AI workflow at 10,000 runs a month is €60 on n8n Cloud Pro and $733.50 on Zapier Professional, both billed monthly. For a few short automations, Zapier's free and entry tiers can cost less than n8n Cloud Starter.
Can I sell n8n workflows to clients?
You can build and run workflows for your own business on the Community edition. Offering n8n itself as a hosted service to customers is outside the Sustainable Use License and needs an agreement with n8n; building workflows as a consultant that run in the client's own n8n is a different situation from hosting it for them.
