Skip to main content
nexeai

n8n: our review

n8n is a visual automation tool that connects your software and slides an AI step into the middle of the flow. Its self-hosted version makes it one of the few tenable choices when the files processed must not leave your servers.

Cloud or self-hostedFree version, then paid·from €20/month
Go to n8n's website (opens in a new tab)

What we take away from it

What counts in its favor

  • Installs on your own servers: the data processed doesn't leave, only calls to the model do, and you choose which one.
  • Priced per execution rather than per user: opening the tool to a whole team doesn't multiply the bill.
  • The code is readable and modifiable, which avoids depending on a publisher for a specific case.

What counts against it

  • The logic is visual, the content of the nodes isn't: as soon as a case gets specific you need to write JavaScript. Plan for a technical lead.
  • The interface is in English, with no French version.
  • The license isn't open source in the strict sense: it reserves use to the company's internal needs.
  • Per-execution billing punishes badly written loops, and the bill arrives afterward.

The best choice if you already have software to connect and someone able to get hands-on with a node. To avoid if you're looking for a tool everyone can use with no training.

The facts

First paid tier
€20/month — Starter, billed annually
Data hosting
Cloud or self-hosted · source
Data verified on
30 July 2026

What happens to your data

Is your data used to train their models?
No
Subprocessors
n8n doesn't publish any model itself. The subprocessor that matters is the provider of the model you plug into the AI node, and it changes from one node to another: that's where to look, not at n8n.
Retention
Self-hosted, executions stay in your own database. Usage telemetry is, however, sent every six hours by default; it can be turned off.

n8n in detail

n8n is a visual automation tool: you connect your apps on a canvas, set the conditions, and the flow runs on its own. For the past two years, it has mainly served to slip an AI step into the middle of an existing chain — read an incoming email, extract the useful data from it, create the quote, notify the salesperson.

What it actually does

An n8n flow reads left to right. A trigger at the start: an email received, a file uploaded, a form filled in, a time of day. Then nodes follow one another, each doing one thing: query a database, call an AI model, write to a spreadsheet, send a notification. Between nodes, conditions steer the flow according to what was found.

The use case that comes up most often with our clients fits in one sentence: stop retyping information from one piece of software into another. An invoice arrives by email, its data goes into accounting, a line gets created in tracking, and nobody typed anything.

Why it interests us in particular

The self-hosted version changes the nature of the problem for a firm or SME handling customer data. Files don't leave your server; only calls to the model go out, and you choose which one — including a model installed on your own premises, in which case nothing goes out at all. This is the basis of most of the chains we build in automation and AI agent work.

That property is rare. The vast majority of automation tools are online services where the content being processed passes through the publisher, with no other option.

Where it falls short

Three things to know before committing.

The first is the technical level. The canvas is visual, but the content of nodes isn't always: an unusual date format or an API not covered calls for a few lines of JavaScript. Without someone able to write them, the project stalls at the first edge case.

The second is billing. The Starter tier is €20 a month, billed yearly, for 2,500 executions. The counter runs per execution, not per user: that's an advantage for a team, less so for a flow triggered in a loop by mistake. Watch the counters in the first weeks.

The third is the licence. The code is public, but the Sustainable Use License restricts use to your company's own internal needs. Internally, no practical constraint. To resell a product built on top of it, the text needs a close read.

Our recommendation

If you already have four or five pieces of software that don't talk to each other, and someone comfortable with the technical side, n8n offers the best balance between what it enables and what it costs. If your main constraint is that nobody on the team wants to see a line of code, look instead at Zapier, even if that means paying more and accepting that data passes through the publisher.

Who it's for, and who it isn't

Suited for

  • SMEs of 10 to 200 people that already have several software tools and want to stop copying data between them
  • Firms bound by professional secrecy that need documents to stay on their own infrastructure
  • Teams with at least one technical profile, even part-time

Not a fit

  • Organizations with no technical lead: the first specific case will block the project
  • Teams that need a French-language interface
  • One-off, one-time needs, where a dedicated tool will cost less than building a chain

Frequently asked questions

Do you need to know how to code to use n8n?
Not to get started: connecting two apps and triggering an action is done with clicks. But as soon as a case falls outside the box — an unusual date format, a multi-branch condition, an uncovered API call — you need to write a few lines of JavaScript in a node. Plan for a technical lead on the team, or support.
Do my client documents leave my infrastructure?
In self-hosted mode, no: files and executions stay in your database. Only calls to the AI model go out, and you decide which one to call — including a model installed on your own premises, in which case nothing leaves at all. Note that usage telemetry goes out every six hours by default: it doesn't contain your business data, and can be turned off.
Is n8n really open source?
No, not in the strict sense. The code is public and modifiable, but the Sustainable Use License reserves usage to the internal needs of your company or to non-commercial use. In practice, you can use it freely internally; you can't resell a product whose value comes essentially from n8n.
How much does n8n cost for a team of ten people?
The bill doesn't depend on the number of users but on the number of executions. The Starter tier is $20/month billed yearly for 2,500 monthly executions. Ten people on the same flow don't cost ten times more, but a flow triggered in a loop does.
What's the difference with Make or Zapier?
Zapier covers the most apps and requires the least technical skill. Make offers a richer visual canvas. n8n is the only one of the three that installs on your servers, and that's almost always the reason it's chosen.

Alternatives to n8n

Our sources

Prices and hosting terms change often. We recheck them periodically, but the publisher's page remains the only up-to-date reference.

Is n8n the right choice for you?

We measure this tool against your real constraints — budget, privacy, your team's skill level — before you sign an annual subscription.

Book a call