n8n vs Zapier vs Make

Three things about this comparison are wrong almost everywhere it appears, and all three are checkable in about ten minutes.

n8n is not open source. Make no longer bills in operations. And Zapier's app count is routinely quoted low.

Everything below was read from the vendors' own pricing pages, README and licence file on 15 August 2026. Pricing and limits move; the structural differences do not. Check the current numbers before you commit.

The billing model is the decision

This is the part that determines your bill at scale, and it is the part comparisons skim.

Zapier bills per task. From its pricing page: A task is counted whenever Zapier successfully completes a unit of work for you. Failed actions are not counted. Triggers, polling and built-in data tools do not consume tasks. So a workflow with eight actions that runs a thousand times costs roughly eight thousand tasks. Cost scales with workflow complexity.

Make bills in credits. Also from its own pricing page: Each action your scenario performs consumes a certain number of credits — most actions consume 1 credit. Note this is a change — Make historically counted operations, and most comparisons still say so. The practical shape is similar to Zapier's: per-action, so again complexity-sensitive, but the per-unit economics differ and some actions cost more than one credit.

n8n bills per execution on its cloud — one workflow run is one unit regardless of how many nodes it contains. Cost scales with volume, not complexity.

That difference is the whole story for anyone running long workflows at volume. A twenty-node workflow is twenty billable units per run on a per-action model and one on a per-execution model. If your workflows are short and numerous, the models converge. If they are long, the gap is enormous.

You will see specific savings percentages quoted for this. Ignore them and do the arithmetic on your own workflow shape — it is multiplication, and the answer depends entirely on your node count.

Integrations, as advertised by each vendor

• Zapier: 9,000+ apps. The widest catalogue by a large margin, and the main reason to choose it.
• Make: 3,000+ standard apps, plus 350+ AI apps.
• n8n: 1,500+ integrations per its README.

Raw counts matter less than whether your specific tools are covered, and how deeply. A supported integration exposing three actions is not the same as one exposing forty. Check the specific connectors you need rather than the headline number — and note that all three can call an arbitrary HTTP API, so an unsupported service is an inconvenience rather than a blocker.

The licence, stated correctly

n8n is fair-code, distributed under the Sustainable Use License — not an OSI-approved open source licence. Calling it open source is the most common error in this comparison, and the difference is not academic. The licence's own limitations:

You may use or modify the software only for your own internal business purposes or for non-commercial or personal use. You may distribute the software or provide it to others only if you do so free of charge for non-commercial purposes.

So: self-host it for your own business, freely. Build a product on top of it and resell that, and you are outside the licence. Additionally, files with .ee. in the filename require a separate n8n Enterprise Licence, and branches other than master are not licensed at all.

For most teams this is fine — internal business purposes covers the normal case. For anyone planning to embed automation in a commercial product, it is the first thing to read, not the last.

What n8n's model genuinely gives you is self-hosting: your data stays on your infrastructure, which is the deciding factor when the workflows touch regulated or sensitive data. Zapier and Make are cloud-only.

Where AI fits

All three now ship AI features and they are converging fast enough that any specific feature list here will age badly. The durable distinctions:

n8n is the most developed for building actual agents — model-agnostic, with the plumbing to construct a loop with tools rather than just call a model in a step. If you are building an agent rather than adding an AI step to a pipeline, this is the one designed for it.

Zapier is the most accessible: AI as a step in a workflow a non-technical person built, plus agent features across its very large app catalogue.

Make sits between them, with a visual canvas that suits branching logic and an AI assistant for scenario construction.

A caution that applies to all three: an AI step inside an automation platform is still an LLM call, subject to everything in why AI models hallucinate and prompt injection. An automation that reads inbound email, passes it to a model, and lets the model decide what to do next has assembled the lethal trifecta out of drag-and-drop components. The platform will not stop you.

How to choose

Choose Zapier if the people building automations are not engineers, if integration breadth is the binding constraint, and if your workflows are short. It is the most expensive per unit of work and frequently the cheapest overall once you count the engineering time it does not consume.

Choose Make if you need branching, iteration and error handling in a visual tool, and Zapier's linear model is fighting you.

Choose n8n if you have technical people, if your workflows are long enough for per-execution billing to matter, if the data cannot leave your infrastructure, or if you are building genuine agents. Budget for running it — self-hosting is free of licence cost, not free of operations.

Choose none of them if what you actually have is one scheduled script. A cron job and forty lines of code has no per-task pricing, no vendor lock-in, and no visual editor to maintain. Automation platforms earn their keep on breadth of integrations and on letting non-engineers build — if you need neither, you are paying for both.

Before any of this, Decide What to Automate and What to Leave Alone scores candidates on volume, stability and failure cost, and Choose Software or a Vendor Without a Six-Month Evaluation structures the selection. The operations pack covers the surrounding process work.

Sources

All retrieved 15 August 2026:

• Zapier pricing — task-based billing, the definition of a task, failed actions not counted, 9,000+ apps
• Make pricing — credit-based billing, most actions consuming 1 credit, 3,000+ standard apps and 350+ AI apps
• n8n LICENSE.md — the Sustainable Use License, its internal-business-purposes limitation, and the .ee. Enterprise Licence carve-out
• n8n README — fair-code positioning, source-available and self-hostable, 1,500+ integrations