n8n vs Make vs Zapier: Which Automation Platform Should You Use?

If you want the cheapest option for complex, high-volume workflows and you have some technical skill, choose n8n. If you want the widest app coverage and the fastest setup for a non-technical team, choose Zapier. If you want a visual builder with fine-grained control at a moderate price, choose Make. The real difference between them is how each one counts usage, so start there.

Key takeaways

• n8n bills per workflow execution (one full run, any number of steps). Zapier bills per task (each successful action). Make bills per credit (each module action).
• That billing unit matters more than the headline price. A long workflow can cost many times more on a per-step model.
• n8n is the only one of the three you can self-host for free with the Community Edition.
• Zapier has the largest directory (9,000+ apps) and the gentlest learning curve.
• All three now ship AI agent and MCP features, but AI steps can consume extra units. Model those costs before you scale.

How each platform charges you

The mistake most buyers make is comparing plan prices without comparing what a "unit" is. Here is how the three vendors define usage, as of October 2026.

n8n: executions

n8n's pricing page defines an execution as "a single run of your entire workflow," regardless of how many steps it contains or how much data it processes. A 3-step workflow and a 30-step workflow each count as one execution per run. Executions that fail before finishing are not billed, according to n8n's "if it doesn't run, you don't pay" promise.

Cloud plans (billed annually, as of October 2026):

• Starter: €20/month, 2,500 executions, 5 concurrent executions.
• Pro: €50/month, 10,000 executions, up to 50 concurrent executions, workflow history and admin roles.
• Business: €667/month, 40,000 executions. Currently self-hosted only, with SSO, Git version control, multiple environments and queue mode.
• Enterprise: custom pricing.

There is also a free self-hosted Community Edition, and a Startup plan that n8n says gives eligible companies under 20 employees 50% off Business.

Make: credits

Make moved to credits as its billing unit. Each time a module performs an action (read, search, create, update, delete, transform, iterate), it usually uses one credit. Some AI features use more. Routers and error-handler modules are free. The Free plan includes 1,000 credits per month. Paid tiers are Core, Pro, Teams and Enterprise. Make's page renders prices dynamically, so check the current figures on the Make pricing page before you commit.

Two behaviors to know: when you run out of credits your scenarios stop (incoming webhooks queue up to your storage limit), and unused credits expire at the end of the term.

Zapier: tasks

Zapier counts a task each time a Zap successfully completes an action. Triggers and polling do not use tasks, and several built-in tools (Filter, Formatter, Paths, Delay, Looping, Sub-Zap, Digest, Storage, Tables and Forms) are free. Plans as of October 2026:

• Free: $0, 100 tasks/month, two-step Zaps only, 15-minute polling.
• Professional: from $19.99/month, multi-step Zaps, webhooks, 2-minute polling.
• Team: from $69/month, 25 users, shared connections, SAML SSO, 1-minute polling.
• Enterprise: custom.

Annual billing saves up to 33%. When you hit your limit, Zapier can switch you to pay-per-task billing at a higher per-task rate, or pause your Zaps if you turn that off. One Zapier MCP tool call uses two tasks.

Side-by-side comparison

A worked cost example

Imagine a lead-routing workflow: a form submission triggers it, then it enriches the lead, classifies it with an LLM, writes to your CRM, posts to Slack and sends a confirmation email. That is roughly five action steps per run. At 2,000 leads a month:

• n8n: 2,000 executions. Fits inside the Starter plan's 2,500.
• Zapier: about 10,000 tasks (5 actions × 2,000), assuming none of the steps are free built-in tools. You would need a higher Professional tier.
• Make: about 10,000 credits plus any extra for AI modules.

None of this includes your LLM bill, which you pay separately to OpenAI, Anthropic or another provider unless you use a platform's built-in AI credits. See our guide on LLM API costs and how to control them to budget that part.

The lesson: count steps × runs before you compare plan prices. For two-step automations, Zapier's simplicity often wins. For long workflows, per-execution billing usually wins.

Where each tool shines

Choose n8n when

• You run long or branching workflows at volume, where per-step billing would get expensive.
• You need data control. Self-hosting keeps workflow data on infrastructure you choose. n8n's hosted plans store data in Frankfurt, Germany.
• Your team is comfortable with JSON, APIs and some code, and wants JavaScript or Python steps.
• You are building AI agents that call tools, with optional human approval before a tool runs.

The trade-off is ownership. Self-hosting means you handle upgrades, backups, uptime and security. If nobody on the team will own that, use n8n Cloud or another tool.

Choose Make when

• You like a visual canvas with routers, iterators and aggregators, and want to see data move between modules.
• You need more logic than Zapier's basic plans without running your own server.
• Your workflows are moderate in length and you can design them to avoid wasted module runs (filters early, routers instead of duplicate scenarios).

Choose Zapier when

• Your team is non-technical and needs automations working this week.
• You depend on niche SaaS apps; Zapier's directory is the largest.
• Most of your automations are short (trigger plus one to three actions), so task counts stay low.
• You want enterprise governance features such as SSO, audit logs and app restrictions on higher tiers.

From experience: In client work across Shopify Plus stores and startup products, I have usually started teams on whichever tool their least technical operator can maintain, then moved only the expensive, high-volume flows to n8n. Migrating a handful of heavy workflows is cheaper than forcing everyone onto a tool they cannot debug.

How to choose in five steps

1. List your top ten automations. For each, write the trigger, the number of action steps and expected monthly runs.
2. Calculate units per platform. Executions for n8n, credits for Make, tasks for Zapier (exclude Zapier's free built-in steps).
3. Check app coverage. Confirm that every app you need has a native integration, or that you can reach it by HTTP request or webhook.
4. Decide who will maintain it. If the answer is "an ops manager with no coding background," weight Zapier or Make. If it is "an engineer," n8n becomes attractive.
5. Run a two-week pilot. Build your most complex workflow on your top choice using the free plan or trial. Watch real usage numbers, error handling and how long debugging takes.

Security and governance checklist

Automation platforms hold API keys to your CRM, email, payments and databases. Treat them as production systems.

• [ ] Use a shared company account, not someone's personal login.
• [ ] Turn on two-factor authentication for every user.
• [ ] Give each connection the narrowest scope the integration allows (read-only where possible).
• [ ] Store credentials in the platform's encrypted credential store, never in plain-text fields or code steps. See API keys and secrets management for startups.
• [ ] Set up error notifications or error workflows so failures do not go unnoticed.
• [ ] Document each workflow: owner, purpose, apps touched, and what breaks if it stops.
• [ ] For AI steps, add human approval before any action that sends money, deletes data or messages customers. Our guide to AI agent security and guardrails covers this in depth.
• [ ] Review usage monthly against your plan limits to avoid surprise overages or paused workflows.

Common mistakes

Comparing headline prices only. A cheaper plan with a per-step unit can cost more than a pricier plan with a per-run unit.

Polling when a webhook exists. Polling adds delay (15 minutes on Zapier's Free plan) and, on some platforms, wasted runs. Use instant webhook triggers where the source app supports them.

Putting filters late. Filter as early as possible so unneeded runs stop before they consume billable steps.

Self-hosting without a plan. A self-hosted n8n instance with no backups, monitoring or update schedule is a liability.

Letting AI act unsupervised. An LLM step that can email customers or edit records needs guardrails and logging from day one.

Next steps

• Write down your ten most important automations with steps and monthly volume, then price them on all three platforms.
• Start a free plan or trial on your top pick and rebuild your single most complex workflow.
• Browse AI automation ideas for small business to find high-value workflows to build first.
• If you are deciding between building custom automations and buying a platform, read build vs buy software decisions.
• Need help designing an automation stack for your team? A fractional CTO engagement can scope it, build the first workflows and hand over documentation.

Sources

• n8n Plans and Pricing: https://n8n.io/pricing/
• Make Pricing and credits FAQ: https://www.make.com/en/pricing
• Zapier Plans and Pricing: https://zapier.com/pricing
• OpenAI API pricing: https://developers.openai.com/api/docs/pricing
• OWASP Top 10 for LLM Applications 2025: https://genai.owasp.org/llm-top-10/

Keep reading