n8n vs Zapier vs Make: Which We Pick for Client Projects

By Haseeb Asad, founder of Codex Labs. Published . Facts checked on 26 September 2026.

For client projects we pick n8n by default: it bills per workflow run instead of per step, self-hosts for free, and handles AI agents with memory and database tools. We recommend Zapier when a non-technical team must own simple automations across many SaaS apps, and Make when visual multi-step scenarios at a low price matter most.

We are a small studio, and most of our automation work is WhatsApp AI agents built in n8n. That is a bias, so here is the reasoning behind it and the cases where we would tell you to use something else. All prices and limits below were checked on each vendor's own page on September 26, 2026.

n8n vs Zapier vs Make: what is the difference?

n8nZapierMake
What you pay forExecutions: one full workflow runTasks: each successful action stepCredits: each module action
Free optionSelf-hosted Community Edition; Cloud plans start with a free trialFree plan, 100 tasks per month, two-step ZapsFree plan, 1,000 credits per month, 2 active scenarios
Entry paid planStarter: 20 EUR per month billed annually (24 EUR monthly), 2,500 executionsProfessional: from $19.99 per month billed annually ($29.99 monthly), 750 tasksCore: from $9 per month for 10,000 credits
Self-hostingYesNoNo
AppsLarge catalog plus community nodes and any HTTP API9,000+ apps, by Zapier's count3,000+ apps, by Make's count
AI agentsLangChain-based agent, memory and tool nodesAgents, a separate product with its own quotaAI modules billed in credits
Best fitDevelopers, long or AI-heavy workflowsNon-technical teams, simple linear automationsVisual builders, multi-step scenarios on a budget

How does pricing compare for real workflows?

The billing unit matters more than the sticker price.

  • n8n counts an execution as "a single run of your entire workflow. It doesn't matter how many steps are in the workflow or how much data it processes" (n8n pricing).
  • Zapier says "a task is any successful action that runs in Zapier." Triggers never use tasks, and filters, paths, formatter and delay steps do not count either (Zapier help).
  • Make replaced operations with credits on August 27, 2025, at a 1:1 conversion. Each module action uses a credit, and some AI modules use more (Make help).

Here is why that decides it for us. In our restaurant WhatsApp agent, a single text message passes through about 20 nodes on its main path: filters, a state lookup, a customer record, a router agent, the main agent, a JSON parser, a switch and the send. On n8n that is one execution. On Zapier or Make, every action step in that path would be billed, so each customer message would cost many tasks or credits. The exact number depends on how you rebuild the flow, but the direction does not change.

For a five-step automation that runs 200 times a month, none of this matters. All three are cheap, and ease of use should decide.

Current plan prices, with the billing basis each vendor shows:

PlanPriceIncluded
n8n Starter20 EUR per month annual, 24 EUR monthly2,500 executions, 5 concurrent
n8n Pro50 EUR per month annual, 60 EUR monthly10,000 executions, 20 concurrent
n8n Business667 EUR per month annual, 800 EUR monthly40,000 executions, self-hosted only
Zapier Professionalfrom $19.99 per month annual, $29.99 monthly750 tasks, multi-step Zaps, 2 minute polling
Zapier Teamfrom $69 per month annual, $103.50 monthly2,000 tasks, 25 users, 1 minute polling
Make Corefrom $9 per month10,000 credits, 1 minute interval
Make Profrom $16 per month10,000 credits, priority execution
Make Teamsfrom $29 per month10,000 credits, team roles

Make's page lets you pay monthly or annually and says annual billing saves 15% or more, so check which period the price you see applies to. Enterprise plans on all three are quoted on request.

Why do we pick n8n for client projects?

Four reasons:

  1. AI agents are first-class. n8n's AI nodes implement LangChain's JavaScript framework (n8n docs). An agent node takes a model, a memory and tools, and a tool can be a Postgres query, a vector search or another workflow. Our WhatsApp agents use all three. See the build log.
  2. Per-run billing and free self-hosting. Long workflows do not get more expensive per step. Our client agents run on self-hosted n8n.
  3. Community nodes and raw HTTP. Our WhatsApp builds use a community node for Evolution API. n8n's docs note that unverified community nodes "require self-hosting n8n", which is one more reason we self-host.
  4. Code where it helps. Models sometimes return their JSON twice or wrapped in code fences. In our restaurant agent a short Code node strips the fences and parses only the first complete JSON block, which no visual step could do cleanly.

The license matters for agencies. n8n's Sustainable Use License allows building automations for clients on your instance "as long as your clients do not have the possibility to create or edit them", and charging for "workflow creation, setup, and maintenance services." It does not allow hosting n8n as a service where clients build their own workflows (license FAQ). If a client wants to build their own workflows, they get their own n8n.

When is Zapier the better choice?

When the people who will own the automation are not technical, and the job is connecting SaaS apps in a straight line. Zapier's catalog is the largest of the three (9,000+ apps by its own count), its builder is the easiest to hand over, and no one has to run a server.

The trade-off is cost at volume. A busy multi-step Zap burns tasks quickly, and Zapier's AI Agents product has its own separate quota.

If a client's team already lives in Zapier and needs a few more Zaps, we would not move them to n8n just because we prefer it.

When is Make the better choice?

When someone on the client's team likes building visually and needs branches, iterators and routers without writing code, on a small budget. Make's paid plans start low, its scenario editor shows data flowing between modules, and paid plans run scenarios as often as every minute.

Watch two things. Every module action is a credit, so loops over large lists add up. And AI modules can use more than one credit each.

Which one handles WhatsApp and AI agents best?

All three connect to Meta's official WhatsApp Business Cloud API: n8n with a native node and trigger, Zapier with its WhatsApp Business app, and Make with its WhatsApp Business Cloud app.

The difference shows up once the agent needs memory per customer, database lookups, a human handoff with its own state, and voice notes. In n8n that is one workflow you can read top to bottom. We would not try to build our restaurant agents on Zapier or Make. For how the connection choice itself works, see Evolution API vs WhatsApp Cloud API. For budgets, see what a WhatsApp automation project costs.

Frequently asked questions

Is n8n cheaper than Zapier?

Usually, for multi-step work. n8n bills per execution, a whole workflow run however many steps it has. Zapier bills per task, each successful action step. A long workflow that runs often costs far less on n8n, and self-hosted n8n Community Edition is free apart from your server.

Can I self-host n8n for free?

Yes. n8n's Community Edition is free to self-host under the Sustainable Use License. You can use it for your own business and build automations for clients on your instance, but you cannot host n8n as a service where clients build their own workflows.

Which is best for AI agents: n8n, Zapier or Make?

For agents with memory, database tools and custom logic we use n8n. Its AI nodes are built on LangChain and connect models, memory and tools on one canvas. Zapier sells Agents as a separate product with its own quota, and Make bills AI modules in credits.

Is Zapier or Make better for non-technical teams?

Zapier is the easiest for a non-technical team to own: it has the largest app catalog, 9,000+ apps by its own count, and simple linear Zaps. Make is visual and cheaper per step but asks more of the person building. n8n suits teams with a developer or a partner.

What counts as a task, an operation or an execution?

On Zapier a task is any successful action; triggers and filters do not use tasks. On Make each module action uses a credit, which replaced operations in August 2025. On n8n an execution is one run of a whole workflow regardless of its steps.

Do n8n, Zapier and Make support WhatsApp?

All three connect to Meta's official WhatsApp Business Cloud API: n8n with a native node and trigger, Zapier with its WhatsApp Business app, and Make with its WhatsApp Business Cloud app. For unofficial gateways such as Evolution API, n8n's community nodes or HTTP requests are the practical route.

Prices and limits checked on n8n.io/pricing, zapier.com/pricing and make.com/en/pricing on September 26, 2026. Vendors change plans often; check their pages before you buy.

Related reading