Zapier vs n8n: The Billing Unit Decides This One
Two automation platforms, two completely different meters. One charges you per successful action, the other per workflow run — and that single choice sets your ceiling.
- Zapier meters a task per successful action step. n8n meters one execution per run of the entire workflow, no matter how many nodes are in it.
- Zapier's meter is more forgiving than people think: triggers never use tasks, and Filter, Paths, Formatter, Delay, Looping, Digest, and Storage steps don't either.
- n8n gives you unlimited users and workflows on every cloud plan and a self-hostable Community Edition — but it is fair-code, not open source, and you can't host it and charge people to use it.
- Blow through Zapier's task limit and the rest of your tasks are held until the billing cycle resets unless you upgrade or turn on pay-per-task.
Zapier bills a task for every successful action step, while n8n bills one execution per run of an entire workflow regardless of node count. Short, high-volume automations are cheaper on Zapier. Long, multi-step AI pipelines are cheaper on n8n, which also self-hosts. That meter difference drives the choice.
I have run both of these in production, and the argument was never about node counts or which canvas looks cleaner.
It’s about what each one puts on the meter.
Zapier charges you per successful action. n8n charges you per workflow run. Everything else in this comparison falls out of that one line.
What each platform actually meters
Zapier counts a task every time it successfully completes a unit of work. What surprised me is how much doesn’t count. Per Zapier’s own task documentation, triggers never use tasks. Neither do Filter or Paths steps. Neither do Formatter, Delay, Looping, Digest, or Storage. And action steps that error or halt are free.
So the “you pay for every step” caricature is lazy. Zapier bills successful action steps, and a real chunk of the plumbing rides along free.
n8n doesn’t count steps at all. Its pricing page puts it plainly: an execution is a single run of your entire workflow, and it doesn’t matter how many steps are in the workflow or how much data it processes.
Three nodes or thirty-three. Same one execution.
The comparison table
| Zapier | n8n | |
|---|---|---|
| Billing unit | One successful action step (a “task”) | One run of the entire workflow |
| Not billed | Triggers, Filter/Paths, Formatter, Delay, Looping, Digest, Storage, failed steps | Extra nodes and data volume inside a run |
| Free option | Free plan: 100 tasks/mo, two-step Zaps, unlimited Zaps/Tables/Forms | Self-hosted Community Edition |
| Users & workflows | Scale with plan tier | Unlimited users and workflows on all plans |
| Concurrency | Managed for you | A plan lever — 5 concurrent runs on Starter, 200+ at Enterprise |
| Self-hosting | No | Yes |
| License | Proprietary SaaS | Sustainable Use License (fair-code, not open source) |
| Ops you own | None | All of it, if you self-host |
Where Zapier wins
Nothing to stand up, nothing to patch, nothing to page you at 2am. That’s the whole pitch and it’s a real one — I came up in Ops, and the cheapest incident is the one that can’t happen on your box.
The forgiving meter matters too. A Zap with a trigger, two filters, a Formatter step and one write costs you exactly one task. For short automations at high volume, that’s hard to beat, which is why it keeps showing up in every honest list of AI tools for small business.
Where it stings: the Free plan caps you at two-step Zaps, so the moment you need real branching you’re paying. And per Zapier’s docs, blow past your task limit and the rest get held until the cycle resets unless you upgrade or switch on pay-per-task. And MCP calls hit the same bucket at a worse rate: Zapier’s docs put every successful tool call at two tasks, with failed calls free.
Where n8n wins
Fan-out. If one run parses a document, enriches it, calls a model, branches three ways and writes to two systems, n8n calls that one execution. That shape is exactly what an AI agent pipeline looks like in practice.
Unlimited users and workflows on every cloud plan means you stop rationing automations, and self-hosting keeps customer data on infrastructure you control.
Where it stings: concurrency is a plan lever, not a given — Starter runs five at a time. And read the license before you build a business on it. n8n is fair-code, not open source, by its own admission. Internal use is fine. Hosting it and charging people for access is not.
Where I land
Short workflows, no ops appetite, connectors are the point… Zapier.
Long runs, multi-model chains, data you can’t hand to a SaaS… n8n.
Stop shopping for the better platform. Start counting what your average run actually does, then pick the meter that doesn’t punish it.
#TheAIMogul
Bottom linePick Zapier when the workflow is short, the connectors matter more than the architecture, and you want zero ops. Pick n8n when your runs fan out into many steps, you're chaining models, or you need the box on your own infrastructure. The billing unit is the decision.