n8n, Make or Zapier: which one fits your business

Zapier starts free with 100 tasks per month, and paid plans start at $ 19.99 per month. Make starts free with 1,000 credits. n8n is free if you host it yourself, and in the cloud you pay from around € 20 per month for 2,500 executions. But the billing model says more than the entry price.

The prices in this article exclude VAT and were checked in October 2026 through several price overviews. Check the current price with the provider.

Zapier: per task

Every action step that runs is a task. A Zap that reads an email, looks something up in your CRM, adds a row to a spreadsheet and sends a Slack message therefore costs several tasks each time. The free plan gives 100 tasks per month for simple Zaps with one trigger and one action. Professional starts at $ 19.99 per month for 750 tasks and Team at $ 69 per month, both with annual billing.

Make: per credit

Until August 2025, Make charged in operations, and since then in credits. Every module that runs costs a credit; AI functions and code cost more. The free plan gives 1,000 credits per month and paid plans start at around $ 9 per month with annual billing.

n8n: per execution

n8n counts one execution each time the whole workflow runs; how many steps are in it doesn't matter. The Community Edition is free if you put it on your own server, with no limit on the number of executions. You then only pay for the server and the management. In n8n's cloud, Starter costs around € 20 per month for 2,500 executions and Pro around € 50 for 10,000, with annual billing.

A worked example

Take a workflow of ten steps that runs 1,000 times a month. In Zapier that's around 10,000 tasks, in n8n 1,000 executions (virtua.cloud, 2026). In Make you count per module, so there too it heads towards 10,000 credits. For a simple two-step integration it makes little difference. For a process with many steps and lots of volume, the difference is large.

Don't forget the hidden costs: the time to build it, the AI usage of the language model (you pay that separately to OpenAI, Anthropic or Microsoft) and the management. A workflow that nobody maintains will break at some point.

Only n8n can be self-hosted. Make and Zapier always run in the provider's cloud. Self-hosting gives you control over your data and low costs per execution, but requires management: updates, backups, security and monitoring.

Self-hosting sounds more technical than it is, and at the same time people underestimate it. The installation is done in no time. The work is in what comes after:

• Updates. n8n releases new versions regularly. You want to keep up, but not update blindly on a Friday afternoon.
• Backups. Your workflows and credentials are on that server. No backup means everything is gone.
• Security. That server has access to your ERP, CRM and mailbox. That calls for good passwords, encryption and a server that isn't open to the whole world.
• Monitoring. You want to know a workflow is failing before a customer calls.

The Community Edition is free to use for your own processes. If you want to resell n8n or offer it as a service to others, read the license terms first. For larger organizations there are paid versions with extra management features such as single sign-on.

No appetite for server management? Then choose n8n Cloud, or have it hosted by a party that manages it for you. At NinA, n8n runs for clients in the EU, with monitoring and maintenance in a fixed contract.

If personal data passes through your workflows, you need a data processing agreement with the provider and you want to know where the data is stored. With n8n on your own server in the EU, you have the most control over that. But note: the AI model you call also processes that data.

A workflow tool sees everything that passes through it: customer names, addresses, order details, sometimes medical or financial information. Three things to arrange:

1. 01Data processing agreement. With the provider of the workflow tool, and with every service you call. With Make and Zapier your data is in their cloud; check in their data processing agreement where and under what conditions.
2. 02The AI model. Hosting n8n yourself in Amsterdam and then sending everything to a model in the US doesn't keep your data in the EU. Choose a business plan that doesn't train on your data, and where needed a model that runs in the EU, such as Azure OpenAI in a European region.
3. 03Execution logs. Workflow tools often keep a copy of what happened during each execution, which is useful when tracking down errors. But that's personal data too. Set how long it's kept.

More on the risks of AI services and what to arrange for each is in using AI safely.

All three have AI steps and some form of agents. n8n goes furthest: you connect any language model, give an agent its own tools and memory, and build checks and an approval step around it. Zapier and Make are easier, but you have less control.

An AI step in a workflow is often enough already: read this PDF and give me the order number, the items and the quantities. All three can do that.

An AI agent goes a step further. It gets a goal and chooses for itself which tools it uses: looking up the order, checking stock, writing a draft reply. In n8n you build such an agent as part of a workflow. You choose the model yourself, whether that's GPT, Claude, Gemini or a model in your own Azure environment. You give it only the tools it needs, and you put fixed steps around it: check first, only then send.

That last point is why we almost always build agents in n8n. Letting an agent work on its own is a gamble. An agent within a workflow with fixed boundaries, logging and a human who approves is simply work that runs every day.

Three questions to assess the tools:

• Can you choose the model yourself? And switch when a better or cheaper model comes out? Our benchmarks show that this can make a considerable difference per task.
• Can you see what the agent did? Which tools it called, with what data, and why.
• Can you put a human in between? An approval step before anything goes to a customer or into your ERP.

Choose Zapier for quick, simple integrations that someone without technical knowledge builds themselves. Choose Make for visual flows with many steps at a low price. Choose n8n for business-critical processes, AI agents, high volumes and data under your own control.

Zapier fits when

• You want a few simple integrations, such as a form to your CRM or a new lead to Slack
• The people building it don't have a technical background
• The volume is low and no sensitive data passes through it

Make fits when

• You want flows with many steps, filters and branches, and want to see it visually
• The budget is small and the volume is average
• You have no requirements on where the software runs

n8n fits when

• It's a core process, such as orders, invoices or customer questions
• There are many steps and lots of volume, and per-step billing gets expensive
• You want to build AI agents with your own choice of model, tools and checks
• You want to keep your data on a server in the EU or in your own environment
• There's someone to manage it: your own IT or a partner

None of the three

Sometimes a workflow tool isn't enough. If employees need their own screen, the AI has to search thousands of documents or it's about very high volume with strict speed requirements, you build custom software. In our APO method that's called build layer 4. And sometimes there's simply standard software that already does it: then buy that.

It rarely goes wrong on the technology. It goes wrong when nobody owns it, when there's no error handling and when the costs quietly add up. Arrange that in advance, whichever tool you choose.

The classic: someone in marketing builds a handy Zap in 2023, with their own account. Two years later they've left, the account expires and suddenly no more leads come into the CRM. Nobody knew it existed.

How to prevent that:

• One business account, not ten personal accounts. Managed by someone who's staying.
• An owner per workflow. Who gets called when it fails?
• Error handling. What happens when the ERP is briefly unreachable? Does the workflow try again, or does the order disappear into thin air?
• A cost ceiling. A workflow that ends up in a loop because of an error can use up your monthly budget for tasks or AI usage in a single night.
• Documentation. One page per workflow: what does it do, which systems, which credentials, who owns it.
• A human in the right place. Not at every step, but where a mistake costs money or affects a customer.

n8n is our standard for workflows and AI agents. We build within the AI Partnership, host in the EU and hand everything over: the workflows, the data and the documentation are yours.

We don't choose n8n because it's trendy, but because it fits how we work. We can choose the model per task, put a human in the right place, log everything and host it where the client wants: on the client's own infrastructure or on servers with a European provider. We work to ISO 27001 and NEN 7510.

This is how it goes:

1. 01AI Quick Scan. One day on site. We map out the processes and calculate per opportunity what automating delivers. € 6,000, with a business case and roadmap within a week.
2. 02Building in the AI Partnership. A fixed team builds the workflows and agents, connected to your systems, such as Exact, AFAS, SAP, Microsoft Dynamics, HubSpot or Shopify. From € 4,500 per month for 30 hours, for at least twelve months.
3. 03Keeping it running. Hosting, monitoring and maintenance through a contract of € 245, € 445 or € 895 per month, including AI usage.

All amounts exclude VAT. No lock-in: if the collaboration ends, you take everything with you within 30 days.

Want to get started yourself? Our freebies page has free n8n workflows to start with. More about what we build with n8n is on n8n automation and AI automation.