n8n Review: One Execution, However Many Steps

Zapier bills per task. Make bills per operation. Both mean roughly the same thing: every individual step in your automation costs you.

n8n bills per execution. One workflow run is one execution, whether it contains three steps or thirty.

Take a realistic directory pipeline. A listing gets submitted. You check whether the domain already exists, look up the business, fetch a logo, categorise it with an AI call, write it to your database, notify yourself, and send the submitter a confirmation. Call it a dozen steps.

On a per-task tool that submission costs you twelve tasks. On n8n it costs you one execution.

If your automations are short, that difference does not matter. Directory automations are not short.

What n8n is

A visual workflow builder. You drag nodes onto a canvas and connect them, and the data flows left to right through the chain.

It has hundreds of integrations, which covers the things a directory touches: databases, spreadsheets, email, Slack, Stripe, HTTP requests to anything with an API. Where there is no dedicated node, there is a generic HTTP node, which in practice means everything is reachable.

The part that separates it from the pure no-code tools is the escape hatch. Any node can run a snippet of JavaScript or Python. So when you hit the thing the visual builder cannot express, you write four lines instead of abandoning the workflow.

For a directory that matters, because data cleaning is exactly the kind of problem where you need one weird transformation in the middle of an otherwise simple pipeline.

Pricing

Note the currency. Pricing is in euros, which trips up budgeting if you assume dollars.

The Starter plan at €20 for 2,500 executions is the one to price against. Under the execution model, 2,500 executions is 2,500 complete workflow runs, not 2,500 individual steps. Compare that against a per-task competitor by counting the steps in your actual workflow, not by comparing the headline numbers.

The self-hosted Community edition is free with unlimited executions and unlimited workflows. You supply the server.

The licence, stated accurately

This is where n8n gets misdescribed constantly, including by people recommending it.

n8n is not OSI open source. It ships under the Sustainable Use License, which the project calls fair-code.

Practically, that means you can use it freely for internal business purposes, including commercially, and you can modify it. What you cannot do is sell n8n itself, or offer it as a hosted service to others, without a commercial agreement. Files in the repository ending .ee are excluded from the Sustainable Use License entirely and carry a separate enterprise licence.

For running a directory this changes nothing. You are using it internally, which is exactly what the licence permits. But if you are choosing tools partly on licence grounds, do not file n8n next to Scrapy or Baserow. It is a different thing, and the distinction is real.

Where it fits in a directory

The scrape-to-database pipe. Scraper produces rows, n8n picks them up, deduplicates against what you already have, normalises the phone numbers and capitalisation, and writes only genuinely new records into NocoDB or Baserow.

Submission handling. Form fires a webhook, n8n validates the URL is live, checks for duplicates, files it as Pending, and messages you. You approve. n8n publishes.

Link rot monitoring. Scheduled workflow walks every listing URL once a week, flags the ones returning 404 or timing out, and gives you a list. This is the maintenance job that keeps a directory from quietly dying, and almost nobody sets it up because doing it by hand is unbearable.

AI enrichment. Because n8n has AI nodes and a code node, you can run each new listing through a model to draft a description or assign a category, then write the result back. That is a multi-step operation, which is exactly where the execution-based billing pays off.

Where it does not fit

You have to self-host to get the free tier, and that is a real server with real maintenance. Cheap in money, not free in attention.

The learning curve is steeper than Zapier's. Zapier is deliberately simple and n8n is deliberately capable, and capability costs clarity. Expect to spend an evening understanding how data passes between nodes before anything complex works.

Five active workflows on the Starter plan is tighter than it sounds. A directory that is properly automated ends up with more than five.

And it is not a scraper. n8n can make HTTP requests and parse responses, but pointing it at a defended site is the wrong tool. Use a real scraper and let n8n move the results.

Common questions

Is n8n open source?

No, and the distinction matters if licences matter to you. It is fair-code under the Sustainable Use License. Free to use internally, including commercially. Not free to resell or host for others.

Does self-hosted have fewer features?

The Community edition is broadly complete for this use. The enterprise-licensed .ee files cover things like advanced permissions and SSO, which is the usual line.

How does it compare to Make?

Make is easier to start and bills per operation. n8n is more capable, bills per execution, and can be self-hosted for free. For short workflows Make often wins. For long ones n8n wins on cost, and the gap widens with every step you add.

Can it call an AI model?

Yes, there are dedicated AI nodes plus the generic HTTP node for anything not covered. Chaining a model call into a data pipeline is one of its better use cases.

What happens if I run out of executions?

The workflows stop until the period resets. Self-hosted, there is no cap to run out of.

Verdict

n8n is the automation layer for a directory that has grown past a single step.

The reason to choose it over Zapier is arithmetic rather than preference. Count the steps in the workflow you actually need, multiply by how often it runs, and compare that against a per-task allowance. For directory work the numbers usually come out clearly in n8n's favour, and self-hosting removes the meter entirely.

Just be accurate about what you are adopting. Fair-code, not open source. For your purposes that is fine, and it is worth knowing rather than discovering later.