n8n vs Zapier vs Make for business automation in 2026 — and when to write code instead
Three tools dominate business automation, and they charge for three different things: Zapier counts tasks, Make counts operations, n8n counts workflow executions (or nothing at all, self-hosted). That difference, more than any feature list, decides which one is cheap for your workload — and there is a point where none of them is the right answer.
How each one charges
List prices change often — the figures below were compiled in May 2026 and should be checked before you commit. Zapier bills per task (each action a Zap performs), from a free tier of 100 tasks a month to paid plans starting around $20 a month for 750 tasks. Make bills per operation (each module run), from 1,000 free operations to paid plans starting around $10 a month for 10,000 — so a multi-step scenario costs several operations per run but each is cheap. n8n Cloud bills per workflow execution regardless of steps, with plans from roughly $20–30 a month; self-hosted n8n is free to run on your own server (a small VPS at $5–10 a month), with unlimited workflows and executions and your data staying on your infrastructure.
The rule of thumb: short, simple automations are cheapest on Zapier; multi-step scenarios are cheaper on Make; high-volume or data-heavy workflows — and anything with privacy requirements — are cheapest on n8n, especially self-hosted.
When Zapier wins
- You need an integration that only Zapier has (its app catalogue is the largest).
- Non-technical staff will build and maintain the automations themselves.
- Volumes are low and simplicity matters more than cost per task.
When Make wins
- Workflows branch, loop and transform data — the visual canvas is the clearest of the three for complex logic.
- You run many steps per execution and want a low cost per operation.
- You want a hosted tool without the overhead of running a server.
When n8n wins
- Volume is high: one execution can run fifty steps for one unit of billing.
- Data must stay on your infrastructure — client records, health data, government data. Self-hosting is the answer and n8n is built for it.
- Developers are involved: code nodes, custom integrations, version control and environments are first-class.
- AI-heavy flows: n8n's agent and LLM nodes make it a strong backbone for lead automation, document pipelines and assistants. Our lead automation for WhatsApp, Instagram and Facebook runs on exactly this pattern, feeding CRMs like GoHighLevel and HubSpot and booking tools like Cal.com and Calendly.
AI steps and agents in all three
All three now offer AI steps — summarise, classify, extract, draft — and agent-style nodes that call tools. The caution from our agents article applies: scope each action, ground answers in approved content, log everything, and keep a human accountable for actions that touch customers or money. A workflow tool is a fine place to orchestrate an assistant; it is a poor place to hide an unreviewed one.
When to write code instead
Move from a workflow tool to code when any of these are true: the workflow is core to your product rather than around it; you need transactions, retries and idempotency you can reason about; volumes make per-execution pricing irrational; or the logic has outgrown a canvas and nobody can read it any more. The usual destination is a small service on Google Cloud (Cloud Run, Pub/Sub, Scheduler) with the workflow tool kept for the edges. We build both — and we will tell you which one your case needs on a scoping call.
FAQ
Which is cheapest: n8n, Zapier or Make?
It depends on the shape of your workflows. Zapier (per task) is cheapest for short, low-volume automations; Make (per operation) for multi-step scenarios; n8n (per execution, or free self-hosted) for high volume and data-heavy flows. Check current list prices before committing.
Is self-hosted n8n really free?
The software is free to run on your own server; you pay for the server (a small VPS is typically $5–10 a month) and for someone to keep it updated and backed up. Data stays on your infrastructure, which is why we recommend it for client, health and government data.
Can these tools run AI agents?
Yes — all three have AI and agent-style nodes. Use them to orchestrate assistants that are grounded in approved content, with scoped actions, logging and a human accountable for customer- or money-affecting steps.
When should we stop using an automation tool and write code?
When the workflow is core to your product, needs transactional reliability, has volumes that make per-execution pricing irrational, or has become unreadable on a canvas. A small service on Cloud Run with Pub/Sub is the usual next step.
Sources
- PricePulse — Zapier vs Make vs n8n pricing comparison (May 2026)
- 1 Sol UK — Marketing & Growth (lead automation)
- 1 Sol UK — AI agents for customer service in 2026