Disclosure: some links in this review are affiliate links and we may earn a commission at no extra cost to you. It does not change the verdict. Learn more.
n8n is the automation tool people move to when Make starts getting expensive. It bills per workflow run instead of per step, it can be self-hosted for free with unlimited executions, and it gives you real JavaScript and Python. It is also the harder of the two to learn, and there is no free plan that lasts. Here is exactly when it is the right choice and when it is not.
The short version
Verdict: n8n is the automation tool that gets cheaper as your workflows get bigger. It charges per workflow run, not per step, so a 30-step automation costs exactly the same as a 3-step one. And you can run it on your own server for free, with unlimited executions.
The catch: it is the harder of the two mainstream options to learn. On G2, “learning curve” appears in 39 reviews and “difficult learning” in another 23 — against 7 for Make. There is no permanent free cloud plan either: 14 days, then €20/month.
Who it is for: anyone with long or frequent workflows, freelancers who need unlimited users, and anyone technical enough to run a small server.
Who should skip it: complete beginners who have never touched an API, and anyone who wants a free plan that does not expire.
What n8n actually is
n8n is a workflow automation tool. Same basic idea as Make or Zapier: a trigger, then nodes connected on a canvas that move and transform data. The differences are structural rather than cosmetic, and there are three of them.
It bills per execution — one run of an entire workflow. In n8n’s own words, “it doesn’t matter how many steps are in the workflow or how much data it processes.”
It can be self-hosted, free, with unlimited executions.
And it shows you the data. Every node exposes the actual JSON going in and out. That is why power users like it and why beginners bounce off it.
Pricing
| Plan | Price (annual billing) | Executions/mo | Concurrent executions | Users | Self-host |
|---|---|---|---|---|---|
| Starter | €20/mo | 2,500 | 5 | Unlimited | No |
| Pro | €50/mo | 10,000 | 20 | Unlimited | No |
| Business | €667/mo | 40,000 | — | Unlimited | Yes |
| Enterprise | Custom | Custom | 200+ | Unlimited | Yes |
Active workflows are unlimited on every plan, including Starter. So are users — which is unusual, and matters if you ever work with a collaborator or a VA.
Prices checked on September 2, 2026 on n8n.io/pricing. The page shows annual billing with a stated 17% saving; the monthly (non-annual) prices are not published, so treat any monthly figure you see elsewhere as an estimate. Overage on Business and Enterprise is billed at €4,000 per additional block of 300,000 executions, invoiced 45 days later.
There is no permanent free cloud tier — only a 14-day trial with Pro features. That is the single biggest difference from Make in the first week of use.
Where the per-execution model wins
This is the whole argument for n8n, so here is the arithmetic.
A 5-step workflow run 2,000 times a month:
- n8n: 2,000 executions. Comfortably inside Starter’s 2,500 at €20.
- Make: 2,000 × 5 = 10,000 credits. That is the entire Core allowance at $12, with nothing left over.
Now make the workflow 20 steps instead of 5:
- n8n: still 2,000 executions. Still €20. Nothing changes.
- Make: 40,000 credits, which is several tiers up the volume slider.
And a workflow running every 5 minutes — about 8,640 runs a month:
- n8n Pro at €50 covers it with headroom, whatever the step count.
- n8n self-hosted covers it for the cost of a small VPS.
- Make needs a large credit tier no matter how short the workflow is.
The rule: short and occasional favours Make. Long or frequent favours n8n. Work out steps × runs before you choose, because the gap between the two is not 20% — at the wrong shape of usage it is an order of magnitude.
Self-hosting: what is actually free, and what is not
This is the part most reviews get wrong, so it is worth being precise.
The Community Edition is free and includes almost the complete feature set, with unlimited executions. It runs under the Sustainable Use License, which permits use “for your own internal business purposes or for non-commercial or personal use,” and lets you share it free of charge for non-commercial purposes. You may not resell it or run it as a paid service for other people. Registering your free Community instance adds folders, debug-in-editor and custom execution data.
What stays paid even when you self-host: custom variables, environments, external secrets, SSO, projects and workflow sharing, Git version control, multi-main mode and log streaming.
The real cost is your time. A small VPS in the €4–10/month range handles a solo workload (n8n does not publish a minimum specification, so treat that as a practical estimate rather than an official one). You own the updates, the backups and the security. If keeping software patched sounds like a chore rather than a Saturday project, pay for Cloud and skip this section.
For anyone handling client data or anything sensitive, self-hosting is the only option here that keeps everything on infrastructure you control.
The learning curve, quantified
Both n8n and Make score 8.4 out of 10 for ease of use on G2, which makes them look identical. The review tags do not.
For n8n, “learning curve” appears in 39 reviews and “difficult learning” in another 23. For Make, learning curve appears 7 times. On Capterra, n8n scores 4.0 for ease of use against Make’s 4.3, and one reviewer states it plainly: the learning curve “is very steep for beginners, especially if they don’t have prior knowledge of APIs.”
That is not a defect, it is a design choice. n8n hands you the raw data and the HTTP Request node — the feature its power users praise most — and expects you to know what to do with a JSON response. If the words “API” and “JSON” mean nothing to you yet, you will spend your first evening confused rather than productive.
Honest recommendation: if this is your first automation tool, learn the concepts on Make’s free plan first, then come to n8n when your workflows outgrow per-step pricing. The concepts transfer directly.
Integrations, templates and code
About 2,100 integrations — roughly 1,500 fewer than Make. Mainstream apps are covered; niche and regional tools more often need a generic HTTP request, which is exactly the thing beginners find hard.
Templates go the other way: around 11,960 community workflows against Make’s 8,250. They skew technical and are contributed by users rather than curated, so quality varies — but for anything unusual, someone has probably already built it.
The Code node runs JavaScript and native Python. On n8n Cloud, Python cannot import external libraries; self-hosted installations can use allowlisted ones. Make, by comparison, gates custom functions behind Enterprise. If you ever want to drop three lines of code into a workflow, n8n is the only one of the two that lets you.
For AI, n8n has AI Agent and LangChain nodes with OpenAI, Gemini and Anthropic support. Wiring an LLM into a workflow is a normal job here, not a project.
Support and reviews
Starter and Pro get the community forum only. Dedicated support with an SLA starts at Enterprise. The GitHub repository has around 203,000 stars, so the community is large and genuinely helpful — but “post and wait” is not a support ticket, and it is worth knowing that before a workflow breaks at an inconvenient moment.
| Source | Score | Reviews |
|---|---|---|
| G2 | 4.7/5 | 298 |
| Capterra | 4.6/5 | 48 |
| Capterra ease of use | 4.0/5 | — |
Recurring complaints: the learning curve above all, and fiddly setup for some integrations — HubSpot and Google Sheets come up repeatedly.
What n8n cannot do
- No permanent free cloud plan. Fourteen days, then you pay or you self-host.
- No official mobile app.
- Git version control is paid-only, even on self-hosted.
- Fewer integrations than Make, so more hand-built HTTP requests.
- No real support below Enterprise.
- Self-hosting is on you — updates, backups, security.
Who should buy which plan
Starter (€20) — the default for one person with a handful of workflows. Unlimited active workflows and unlimited users at this price is the best value in the category, as long as 2,500 executions covers you.
Pro (€50) — anything on a schedule. If a workflow fires every few minutes, 2,500 executions disappears in days and Pro’s 10,000 is what you actually need.
Self-hosted Community (free + server) — high volume, sensitive data, or you already run a server. Unlimited executions for the cost of the VPS.
Business (€667) — not an individual’s plan. Skip it.
How to test it in one weekend
- Start the 14-day trial. n8n plans → You get Pro features, so test at full capability.
- Rebuild one workflow you already understand. Do not learn a new tool and a new problem simultaneously.
- Search the template library before building anything from scratch.
- Open the data panel on every node and read the JSON. This is the skill that makes n8n click; avoid it and you will stay confused.
- Count your executions per month, not your steps. That is your plan.
- Only try self-hosting if you already maintain a server for something else.
Frequently asked questions
Is n8n free?
The self-hosted Community Edition is free with unlimited executions, under a licence that allows personal and internal business use but not reselling it as a service. You pay for the server. n8n Cloud is not free beyond a 14-day trial.
What counts as one execution?
One run of an entire workflow, regardless of how many nodes it contains or how much data passes through. This is the core difference from Make, which charges per step.
Is n8n harder than Make?
Yes, measurably. Learning curve appears in 39 n8n reviews on G2 versus 7 for Make, and n8n scores lower for ease of use on Capterra. Some familiarity with APIs and JSON makes a large difference.
Can I move my Make workflows to n8n?
Not automatically — there is no importer, and you rebuild by hand. The concepts map directly, so a workflow you genuinely understand is straightforward to reproduce. Keep notes on your triggers and field mappings before you start.
Does n8n work with AI models?
Yes — AI Agent and LangChain nodes with OpenAI, Gemini and Anthropic support, plus real JavaScript and Python in the Code node for anything the nodes do not cover.
How we reviewed this
We checked n8n’s official pricing page, documentation and licence file on September 2, 2026, and cross-referenced ratings and recurring complaints on G2 and Capterra. We have flagged in the text every figure n8n does not publish openly, including its monthly (non-annual) prices and any minimum server specification for self-hosting.
Some links in this review are affiliate links: if you subscribe through them we may earn a commission at no extra cost to you. It changes nothing about the verdict — this review sends beginners to a competitor and tells experienced users to self-host for free, neither of which pays us anything. See our Affiliate Disclosure.