A business owner showed me his Zapier invoice. Eighteen euros a month when he started. Three hundred and forty euros eighteen months later. No new automations had been added. Just more customers. That is where an honest n8n vs Zapier comparison has to start.
The billing unit affects usage, but the total cost also depends on set-up, maintenance, infrastructure, licences and external services. Compare those items for the same scope and the same period.
Below is the comparison without the marketing talk. Including the scenario where Zapier is simply the better choice, because that scenario exists.
The difference that decides everything: task versus execution
Zapier counts billable actions for Zaps and activities for Agents. Paid n8n plans use workflow executions; Make charges credits that depend on modules, AI and connection type. Free steps, retries and plan conditions can change the count. So n8n vs Make comes down to the same question as n8n vs Zapier: what exactly is being counted.
With one integration of two steps you will not notice any of this. The difference appears when your automations mature, and they usually do. You start with “form to CRM”. Six months later it is: form → duplicate check → enrich → CRM → email → task → notification → log entry. Eight steps.
Scroll the table sideways to see all columns.
| Example | Count under an explicit assumption |
|---|---|
| 1,000 runs with 8 billable standard actions | 8,000 Zapier tasks, only if all 8 actions per run succeed and are billable |
| 1,000 complete n8n workflow runs | 1,000 executions under the applicable plan; additional services separately |
| Make | Credits depend on modules, AI and connection type |
More billable actions and more runs can together push usage up. How the invoice changes also depends on tiers, free steps, product rules and contract. A quadratic price increase is not a general rule.
The rule of thumb to test your own situation
Estimate for standard actions: successful runs × average number of actually billable actions per run. Check free steps, loops, retries, Agents and the current supplier rules separately.
There is no universal threshold at which switching becomes cheaper. Compare actual usage with the applicable tiers and with the full cost of set-up, maintenance, infrastructure, storage, licences and external services.
n8n vs Zapier vs Make side by side
Scroll the table sideways to see all columns.
| Item | n8n self-hosted | Zapier | Make |
|---|---|---|---|
| Cost model | Infrastructure, maintenance, usage and applicable licences | Tasks for Zaps; Agents has an activity bundle | Credits; modules and AI usage determine consumption |
| Infrastructure | Your chosen infrastructure; assess external services separately | Cloud platform; check current locations and agreements | Cloud platform with an EU or US region per organisation |
| AI agents | Agent workflows with suitable models and tools | Zapier Agents available; assess plan limits | Make AI Agents available; assess model and connection type |
| Switching | Workflows exportable; accounts and licences separate | Budget for rebuilding and checking integrations | Budget for rebuilding and checking integrations |
| Time to go live | Depends on scope, integrations and acceptance test | Depends on scope, integrations and acceptance test | Depends on scope, integrations and acceptance test |
When Zapier or Make is simply better
I build on n8n. That does not make me neutral, so I am explicit about when you do not need me:
- You need one or two simple integrations. A cloud service can then be a good fit. Still check the apps you need, billable actions, limits, external costs and maintenance before you choose.
- Your integration involves a niche app. Zapier has thousands of connectors. If your industry software is among them and has no decent API, you will not win that breadth back.
- Nobody wants to be responsible for a server. Entirely legitimate. Then choose cloud, or have someone else handle the maintenance, but make that choice deliberately.
- Your flows are short and volume stays low. Compare the total cost for your specific use; even at low volume, features, licences and maintenance can decide the choice.
What ‘open source’ does and does not mean for n8n
n8n is often called open source. Strictly speaking, that is not correct. n8n uses the Sustainable Use License, a fair-code licence: you may view the source code, modify it and run it yourself, including commercially within your own business. What you may not do is resell n8n to third parties as a competing hosted service.
The source code is available under the applicable n8n licences; this is not an unrestricted open-source licence. Before use, hosting for third parties, embedding or resale, check which terms and any additional licence apply. See the n8n licence.
When is switching worth it?
Three signals that it is time to price up a Zapier alternative. Two out of three is reason enough to run the numbers:
- There is no universal tipping point. Compare one-off set-up, the chosen maintenance, infrastructure, storage, licences and operational usage over the same period. My maintenance packages start at €250 per month excluding VAT and external costs; at lower volume a cloud service can be cheaper. See the current prices.
- You run into limits you cannot explain. A step you are not allowed to add, or a filter that still costs a task halfway through.
- A client or accountant asks where your data is. Once that question is asked, it is no longer a technical choice but a commercial one.
Switching takes an inventory, rebuilding, authorisations and acceptance tests. Budget for the one-off effort and any temporary double costs. A lower monthly bill is not guaranteed; work it through with the method in what automation costs.
Why I build on n8n
I choose n8n when the combination of workflow logic, infrastructure you choose yourself and AI integrations fits. Paid features, maintenance and external model costs remain separate items; connected services may process data outside the n8n environment.
The full reasoning, including the objections I take seriously, is on why n8n. New to the tool? Start with what is n8n. Is data ownership an issue for you? Then read AI Act, GDPR and your data.
Three years ahead
Comparing single monthly prices is misleading, because the curves run differently. Below is the same business at three points in time. Starting point: a growing SME that begins with three flows and ends with fifteen.
Scroll the table sideways to see all columns.
| Scenario | Runs per month | Billable actions per run (assumption) | Tasks under that assumption |
|---|---|---|---|
| Year 1: 3 flows | 1,500 total | 4 | 6,000 |
| Year 2: 8 flows | 6,000 total | 6 | 36,000 |
| Year 3: 15 flows | 15,000 total | 8 | 120,000 |
There is no well-founded fixed cost line for self-hosting. For each year, work with the capacity needed, the appropriate maintenance, external usage costs and any additional licences.
With self-hosting, too, capacity, storage, licences and maintenance can grow. My standard packages have limits of 3, 5 or 10 named workflows; fifteen workflows do not automatically fall within such a package. Reassess the scope as you grow.
Why there are no exact amounts here
I could have made a table with euros. It would look more convincing and be out of date within six months: all three providers have changed their tiers in the past eighteen months.
Use your invoice and actual usage logs. For each product, count the billable actions, credits or executions, including loops and retries. Simply dividing a task count by all the steps drawn is not a reliable comparison. Check Zapier tasks, Make credits and n8n plans.
What comparison tables leave out
Four practical differences you only notice once you are in the middle of it, and that often weigh more than the price.
Waiting costs money with one and nothing with another
Check how the chosen supplier counts wait steps, polling, retries and failed actions. Not every wait or check step is paid. A count belongs to the specific product and plan, not to a general comparison.
The HTTP node is an escape hatch you will need one day
Thousands of ready-made integrations sounds like an advantage until your industry software is not among them. Then the question is not how many connectors there are, but whether you can build one yourself. With an HTTP node and a code node, you can. Without those two, you are stuck at the first Dutch software package that has no standard connector, and there are many of those.
Testing on real data is an underrated difference
Can you step through a flow with the actual content of a real invoice, and rewind when it goes wrong? When you build a complex chain, that saves hours per flow. Ask for it explicitly in a demo, with your own data.
Ownership of your logic
Your flows are business logic: the recorded agreements on how your business works. The question is whether you can export them, put them under version control and take them with you. In a self-hosted set-up, every flow is a JSON file that you put in your own repository. On a closed platform, it is an export in a format that only works there.
FLOW Migration
What a switch looks like in practice
Who it is for: businesses with five to fifteen existing flows on a per-task platform
Scroll the diagram sideways to see all of it.
What it gives you: No big bang. You run both platforms side by side for two weeks and compare the results before you cancel the old subscription.
Watch out: Only cancel the old subscription after the parallel period. The temptation to save money straight away is strong, and the cost of a missed order is higher.
A working template to look at: n8n template library
An honest step-by-step plan if you want to switch
- Take stock. Which flows are running, how often, and how many steps do they have? This often already shows that a third of them are never used any more.
- Compare actual usage. Use logs to estimate billable actions, credits and executions separately. Free steps, loops and retries make a simple division unreliable.
- Start with your heaviest flow, not your easiest. That is where most of the money is, and you learn straight away whether the switch is realistic.
- Run in parallel for two weeks. Both platforms active, results side by side.
- Only then switch off the old one. And keep the export for another quarter.
With five to ten flows, this takes two to four days of work in practice. Include it as a one-off item in the method from what automation really costs, and you will see straight away whether it adds up.
Frequently asked questions
What is the difference between n8n and Zapier?
Zapier counts billable actions for Zaps; n8n uses workflow executions in its paid plans. Eight thousand tasks for a thousand runs is only an example if every run carries out eight billable standard actions. Self-hosting determines the n8n infrastructure; connected suppliers and data flows are assessed separately.
Is n8n cheaper than Zapier?
That cannot be said in general. Compare one-off set-up, maintenance, hosting, storage, applicable licences and operational usage. Self-hosted does not mean you only pay for the server. A cloud service can be cheaper for your scope.
When is Zapier the better choice?
When you need one or two simple integrations with niche apps, when nobody in your business wants to manage a server, and when you want to be live within an hour. Zapier has thousands of ready-made connectors; you will not win that breadth back with technology.
Can I switch from Zapier or Make to n8n?
Often, yes, if the integrations and features you need are available. Take stock of workflows and accounts, rebuild the logic and test the results and the handover. Budget for migration, temporary double costs and ongoing maintenance in advance; keep the old environment until the agreed acceptance is complete.
Where is my data with n8n, Zapier and Make?
With Zapier and Make you use the supplier’s infrastructure; check current regions, subprocessors and transfer arrangements. With self-hosted n8n you choose the n8n infrastructure, but AI, CRM, email, storage and other services may process data elsewhere. Assess the whole chain.
I build on n8n, so I have an interest in this comparison. That is why the scenarios in which Zapier or Make wins are spelled out. Prices are published rates at the time of writing; check them with the supplier before you decide.