LEETLABS
§ 00 / FIELD NOTE
BY Said Betmurzaev · Founder, Leetlabs

n8n vs Zapier vs custom automation: how to choose

The honest version of the n8n versus Zapier question is a three-way one, because “build it ourselves” is always the third option, and the right answer depends less on features than on how you’re charged, what you’re allowed to do with the tool, and how much the workflow matters to your business.

Zapier: reach, at a per-task price

Zapier’s advantage is coverage. It advertises connectors for more than 9,000 apps, which is more than anything else in the category, and for a huge range of “when X happens, do Y” tasks it’s the fastest path from idea to working automation. Nothing to host, nothing to maintain.

The catch is the pricing model. Zapier charges per task, meaning each action a workflow performs counts against your plan. That’s cheap when volume is low and stays predictable only if your volume does. As automations move from convenience to infrastructure and fire thousands of times a day, per-task pricing turns into a bill that grows with your success, which is exactly when you’d rather it didn’t. Zapier is priced as a convenience, and it’s an excellent one, right up until an automation becomes load-bearing.

n8n: control, and a license worth reading

n8n answers the two things Zapier can’t. First, you can self-host it, and the self-hosted Community Edition runs unlimited executions at no per-task cost, so heavy-volume workflows don’t scale your bill linearly. Second, it’s source-available, so you can see and extend how it works. Its catalog is smaller than Zapier’s (hundreds of integrations rather than thousands), but it’s far more open to custom logic and custom nodes.

The detail people miss is the license. n8n is not classic open source. It ships under the Sustainable Use License, a “fair-code” license that lets you use, self-host, and modify it freely for internal business purposes, but restricts using it to build a competing product or a hosted service whose value comes substantially from n8n itself. For internal automation, where you aren’t reselling the tool, that restriction almost never bites. If your plan is to wrap n8n and sell it, read the license carefully first, because that’s the case it’s written to prevent.

n8n is the right pick when you’ve outgrown per-task pricing, you want automations you can host and control, and your workflows have enough custom logic that a rigid template builder gets in the way.

Custom automation: when the tool becomes the constraint

At some point neither tool fits, and forcing the fit costs more than building. The signals are consistent:

  • The logic is genuinely complex: heavy branching, state that has to persist across steps, real error handling and retries, or idempotency the platform can’t guarantee.
  • The workflow is business-critical and you can’t accept being unable to test it properly, version-control it cleanly, or own the code it runs on.
  • You’re already writing significant custom code inside the tool to make it behave, which means you’ve inherited the platform’s limits without the benefits of a real codebase.
  • Cost, control, or compliance means the automation has to run entirely inside your own infrastructure with no third-party dependency in the path.

A custom build costs more upfront and removes the ceiling. You own the code, you test and deploy it like any other software, and it does exactly what your process needs instead of what the tool allows.

A decision rule that holds up

Reach for Zapier when you want a common integration working today and volume is modest. Move to n8n when per-task pricing starts to hurt or you need self-hosting and custom logic, and read its license if you ever intend to resell rather than just use it. Build custom when the automation is core to how the business runs, the logic has outgrown what a builder can express cleanly, or you need full ownership and control.

The mistake we see most is staying on a per-task tool well past the point it made sense, because migrating feels like effort, while the bill and the workarounds quietly compound. The tool that was cheapest at low volume is rarely cheapest once the workflow becomes infrastructure.

That transition is where our workflow automation work fits: not to talk you out of Zapier or n8n when they’re the right call, but to build the reliable, observable, owned version once a process has become important enough that the convenience tools are costing you more than they save.