n8n vs Make: which one should your business build on?

The short version

Both tools connect apps with visual flows. What decides the bill and the fit is how usage is counted, where the tool runs, and how deep the AI features go. We build most client automations on n8n (HeroStoryBook's WhatsApp support agent and its daily Pinterest publishing run on it), we published 30 free n8n workflows for small businesses, and we work inside Make when a client already uses it.

Pricing side by side

From n8n.io/pricing and make.com/pricing on 19 October 2026, billed yearly unless stated. Make shows visitors from the US a higher price list than visitors from Europe and elsewhere.

Real workflows, priced on both

Across the 30 workflows in our free library, the median has 8 steps and the largest has 15. Here is what different sizes cost each month, billed yearly:

A run processes one item. A trigger that checks for new data every 15 minutes adds about 2,880 credits a month on Make; n8n only counts the checks that find something.

For short workflows Make wins on price. The gap turns as soon as a run works through a list. If each run handles 50 rows from a sheet, Make counts every module once per row, so the same 8-step workflow uses about 400 credits per run. n8n still counts one execution.

Self-hosting and data control

n8n can run on your own server with its free Community Edition, and your customer data then never leaves infrastructure you control. That matters for clinics, law firms and anyone with data-residency rules. The free edition leaves out team features such as SSO, shared projects, environments and Git versioning. n8n Cloud runs in Frankfurt. Make is hosted only: you choose a US or EU data zone, but Make says its own AI assistant, AI web search and OpenAI models always run in the US.

AI agents

n8n's AI Agent node puts a model, memory and tools inside a workflow, with MCP support and an approval step before a tool runs. A whole agent run counts as one execution. HeroStoryBook's support agent reads a WhatsApp message with a voice note and a photo, decides and replies, and each incoming message costs one execution, whatever happens inside. Make's AI Agent has been in open beta since 2 February 2026 and can use modules, scenarios and MCP tools; with Make's own AI provider it charges credits by tokens, while your own API key keeps it at one credit per action.

Ease of use

Make's editor is the friendlier of the two for a first automation, and its templates get a non-technical person to a working scenario faster. n8n exposes more: expressions, code nodes, error handling, versioning, which is what an agency wants and what a beginner may not. If you will build and maintain it yourself, weigh that honestly. If someone builds it for you, the editor matters less than the bill and the data.

When to choose which

• Choose Make if you build automations yourself, they are short, and they run a few thousand times a month or less.
• Choose n8n if workflows are long, frequent or work through lists, an AI agent is involved, or data must stay on your server.
• Weighing Zapier as well? Our n8n vs Zapier comparison prices the same workflows there.
• Already on Make? Stay unless the bill or the AI needs say otherwise. We build inside Make in that case and move scenarios one by one only when it pays.

What we build on n8n, and what it costs, is on our n8n automation agency page. If you would rather run the agent yourself, on your own server and your own model keys, the open-source route is covered in the Hermes Agent guide.

Automating Instagram from either tool? If a post fails with "Container Publication failed", our fix for that error shows how we traced it to file size. For the CRM those automations feed, our Twenty CRM review covers the open-source option we installed ourselves.