Somewhere around the fifth manual copy-paste, a team decides it has had enough. Five jobs get done by hand every week — form responses copied into a sheet, a Slack ping when a payment lands, invoices pushed to accounting — and somebody finally asks why. The search starts with n8n, because a technical colleague has heard the name, and ends with half a dozen pricing pages that each measure something different.
This is not a ranked list. The differences that matter between n8n, Zapier, Make, Activepieces and Windmill in 2026 are not feature checkboxes — they are how each platform counts your usage, where your data is allowed to sit, and who on your team can keep the automations alive. Get those three right and the feature list stops being the deciding factor.
Fix your constraints before you open a pricing page
Three questions split this market cleanly, and none of them is about which tool has more integrations.
The first is who maintains the thing. A developer who is comfortable reading JSON, debugging a webhook and running Docker will be happiest with a tool that exposes those skills. A marketing team that just wants to connect a form to a spreadsheet will not. The same product can be perfect for one and a trap for the other.
The second is where your data may live. Every hosted platform routes data through its own servers while a workflow runs. If you handle patient records, legal documents or anything an EU or industry rule governs, “can we self-host this” becomes the first question. n8n, Activepieces and Windmill all ship a self-hostable edition; hosted-only tools do not.
The third is the billing unit, because it decides what “cheap” even means. Zapier has historically charged per task, Make charges per operation (roughly, per module that runs), n8n charges per workflow execution regardless of how many steps are inside it, and Activepieces charges per credit with pay-as-you-go overage. A workflow that loops over 500 spreadsheet rows is a rounding error on one of these models and a genuine invoice on another.
What each platform is actually for
n8n is the developer’s answer, and it says so. The company markets its tool to “technical teams,” names Microsoft, Meta and Nvidia among its users on its own site, and its GitHub repository passed 206,000 stars by early October 2026. The Community Edition is free and self-hosted; n8n Cloud starts at about $20/month billed annually for 2,500 workflow executions, moving to about $50 for 10,000 and about $800 for 40,000, per n8n.io. Every plan includes unlimited users and unlimited steps per execution, which is the detail that keeps loop-heavy workflows affordable.
Zapier is still the widest on-ramp and the easiest to hand to a non-technical teammate. It is also the one where costs can grow fastest, because its history is task-based billing — every action inside a Zap counts. Zapier no longer publishes a simple public number for its free tier, so check the official pricing page for current limits before you commit a large workflow to it.
Make sits between those two: a visual, node-based canvas a non-developer can learn, billed by operations rather than tasks. Its free tier caps the number of active scenarios and enforces a minimum interval between scheduled runs; paid tiers lift both. Make also now bundles its own AI features — Maia, an MCP server, and AI agents — across plans, per make.com.
Activepieces is the open-source challenger worth knowing in 2026. Its free plan gives 1,000 credits a month with unlimited flows, and its cloud plans — Plus at about $20/month, Team at about $200/month — are flat fees with overage at roughly $0.007 per credit, and it can run on your own servers. If the n8n self-hosting path feels heavy but you still want an escape route from a hosted-only vendor, this is the middle ground.
Windmill is for teams that would rather write Python or TypeScript than drag boxes. Its free and open-source edition runs unlimited executions with up to 50 users, built around scripts, flows and data pipelines rather than a connector catalogue — a different shape of tool, but a fit when the “automation” is really engineering.
The decision, by situation
| Your situation | Start with | Why |
|---|---|---|
| A developer is available and data must stay in-house | n8n Community Edition (self-hosted) | Free, per-execution billing, full control of where data sits |
| A non-technical team needs it working this week | Make or Zapier | Visual builder, largest connector ecosystems, no server to run |
| You want cheap cloud plus an exit route | Activepieces | Flat monthly plans, open source, movable to your own server |
| The automation is really code and data pipelines | Windmill | Script- and pipeline-first, unlimited executions on free |
| Very high-volume, many-step workflows | n8n Cloud | Unlimited steps per execution, so loop-heavy flows stay affordable |
The three ways an automation bill goes wrong
The sticker price is rarely the one that surprises people.
Per-task and per-operation billing multiplies on loops. A workflow that looks like five steps can run 500 times a day if it iterates over a sheet. On task- or operation-based plans, that is where the cost concentrates. Before you pick, estimate how many times the workflow actually fires, not how many boxes sit on the canvas.
Free tiers are not all shaped the same. Some reset daily, some monthly, some cap active workflows rather than runs. n8n counts a workflow once from start to finish no matter how many steps are inside it; a competitor may count every step. Two plans labelled “free” can differ by an order of magnitude for the same job.
Self-hosting is not free — it is a trade. You avoid the subscription, and data leaves your network only if you tell it to, but you now own a server, backups, upgrades and the developer hours to fix a broken credential at 11pm. For one workflow, a $20/month hosted plan is usually cheaper than those hours. For a team already running infrastructure, the maths flips.
For where agent-style automation fits into a small business, see our guide to AI tools for business; if the vocabulary is new, start with AI agents for beginners.
FAQ
Is n8n free? The Community Edition is free and self-hosted; n8n Cloud is a paid subscription starting at about $20/month billed annually for 2,500 executions. Check the official site for current tiers.
Can I move my automations from Zapier to n8n, or back? Not seamlessly. Some tools export workflow JSON, but schemas are not portable between vendors, so treat any migration as a rebuild.
Which of these can I self-host? n8n, Activepieces and Windmill all offer self-hostable or open-source editions. Zapier and Make are hosted-only.
Do these tools have AI features? Yes, and it is the main 2026 development. n8n ships an AI Agent node and MCP support on its paid plans, Make has Maia and an MCP server, and Activepieces bundles agents and MCP tools. These let a workflow call a model or an agent as a step — useful, but review one before trusting it with anything irreversible.
What is cheapest for one person? Any of the free tiers, provided your usage fits their shape. The trap is not the plan price but the overage: on a credit-based plan, a runaway loop can spend a month of credits in an afternoon.
What to do next
Decide by who will maintain the automation and where the data is allowed to live, then pick the platform that matches your billing reality — not the longest feature list. Start on one free tier, rebuild your single most annoying manual task this week, and watch what it actually costs to run. That first workflow will tell you more about the right platform than any comparison table, including this one. When you want to compare more broadly, how to compare AI tools covers the method rather than the products.
Prices and plan details in this guide were checked against vendor sites in October 2026 and change often; confirm current figures on each official pricing page before subscribing.