What is n8n? How it works, self-hosting and your first automations
n8n is an AI-ready automation platform you can self-host. Learn how it works, what it costs, what its licence allows and how to build your first workflows.
Read
The short answer: Zapier is the quickest to set up and has the most integrations, but it gets the most expensive once processes have many steps and run often. Make gives you a lot of power for little money, as long as your scenarios can live in its cloud. n8n wins for complex, high-volume and AI-heavy automation, and whenever data needs to stay on your own server, because it’s the only one of the three you can self-host.
| Feature | n8n | Make | Zapier |
|---|---|---|---|
| Billing unit | Execution (one full workflow run) | Credit (usually 1 per module operation) | Task (1 per successful action) |
| Free option | Self-hosted Community, 14-day cloud trial | Free: 1,000 credits/month | Free: 100 tasks/month, two-step Zaps |
| Self-hosting | Yes | No | No |
| Integrations (vendor figures) | 1,500+ | 3,000+ | 9,000+ |
| Custom code | Code node (JavaScript, Python), custom nodes | HTTP module for any API | Code steps (task usage depends on runtime) |
| AI agents | AI Agent node plus Agent Builder (Preview) | Make AI Agents | Zapier Agents (separate “activities” quota) |
| MCP | Client and server | Make MCP server | Zapier MCP (2 tasks per tool call) |
| Learning curve | Steepest | Moderate | Gentlest |
Integration counts come from each vendor: the n8n README, Make’s comparison page and Zapier’s pricing page. Treat them with a pinch of salt, since all three can reach any service with a public API through a plain HTTP request.
This is the biggest difference between them and the root of most billing surprises. The same process can cost wildly different amounts on each platform.
n8n counts one execution no matter how many steps the workflow has or how much data it moves. The docs add that manual test runs, sub-workflow calls, error workflows and polls that find no new data don’t count. The catch: every incoming webhook is its own execution, and a schedule firing every 5 minutes adds up to roughly 8,600–8,900 executions a month.
According to Make’s pricing page, every action a scenario performs uses credits, usually one per operation: reading, searching, creating, updating, transforming data, iterating over rows and reading from a webhook. The Router and error-handler modules are free. Make itself notes that a single scenario run can use anywhere from two credits to thousands, because looping over 500 rows means 500 operations. AI features are charged on top by tokens if you use Make’s own AI provider (see Make Help Center).
In Zapier, every step that successfully completes an action is a task. Triggers and polling for new data are free, and so are the built-in tools: Filter, Formatter, Paths, Delay, Looping, Sub-Zap, Digest and Storage. Go over your limit and Zapier switches you to pay-per-task billing at a higher rate than your plan, or pauses your Zaps if you’ve turned that option off.
Take a typical flow: a form webhook, a formatting step, a condition, then six app actions (CRM, spreadsheet, email, Slack, a task in your project tool and a database write). It runs 1,000 times a month. The maths below is simplified and assumes no loops and no AI steps:
| Platform | Usage per run | Monthly usage | Fits the cheapest paid plan? |
|---|---|---|---|
| n8n | 1 execution | ~1,000 executions | Yes, Starter includes 2,500 |
| Make | ~8 credits (webhook, formatting, 6 actions; router free) | ~8,000 credits | Yes, Core starts at 10,000 |
| Zapier | ~6 tasks (trigger, Formatter and Filter are free) | ~6,000 tasks | No, Professional starts at 750, so you need a higher tier |
The takeaway: with many steps and high volume, per-execution pricing comes out cheapest. With a handful of simple Zaps a day, the difference stops mattering and convenience wins.
| Plan | n8n | Make | Zapier |
|---|---|---|---|
| Free | Self-hosted Community, no licence fee | 1,000 credits/month | 100 tasks/month |
| Entry | Starter: €20/month, 2,500 executions | Core: ~$9–12/month, 10,000 credits | Professional: from $19.99/month, 750 tasks |
| Mid | Pro: €50/month, from 10,000 executions | Pro: ~$16–21/month, 10,000 credits | — |
| Team | Business: €667/month, 40,000 executions (self-hosted) | Teams: ~$29–38/month, 10,000 credits | Team: from $69/month, up to 25 users |
| Enterprise | Custom | Custom | Custom |
All figures assume annual billing and exclude VAT or sales tax. Monthly billing costs more on all three. n8n prices come from its official pricing page: Starter and Pro are hard-capped with no overage fees, while Business charges €4,000 per extra 300,000 executions. Zapier’s figures come from its pricing guide and pricing page.
Make gets a range because Make itself publishes different numbers. Its Make vs n8n comparison page lists Core at $9, Pro at $16 and Teams at $29, while a Make blog post dated 4 August 2026 quotes “from $12”, $21 and $38 at the 10,000-credit tier. The calculator on make.com/pricing prices each credit tier dynamically, so check the figure for your own volume before buying.
All three now ship AI agents, but they differ in maturity, in how much control you get and, crucially, in how model usage is billed.
If your agent needs to work inside your own systems and data rather than just send emails, see how we build AI agents for businesses.
This criterion often settles the decision on its own. Make’s own comparison page describes it as cloud-first on every plan, and Zapier is cloud-only too. n8n can run on your own server in whichever country you choose, and n8n Cloud stores data in Frankfurt, Germany (n8n pricing FAQ).
Self-hosting doesn’t make you GDPR-compliant by itself, but it simplifies a few things. Customer data doesn’t pass through another processor, you decide how long execution logs are kept, and you can put the editor behind a VPN. With any cloud platform you’ll need a data processing agreement, a check on data location and an entry in your records of processing.
A licensing caveat: you can use free n8n to run automations for clients, but the n8n licence FAQ forbids offering them the workflow editor as a service. If an automation builder for your customers is the product, you need a commercial licence.
Zapier holds your hand: pick a trigger, pick an action, map the fields, done. Its Copilot assistant can draft a Zap from a description. Make shows scenarios as a map of connected modules and handles branching, iteration and aggregation brilliantly, but you need to understand how data “bundles” flow. n8n is closest to programming. You see data as JSON, write expressions in JavaScript, and get full power from the Code node. n8n’s AI assistant can build workflows from a description, but production automations still need someone technical to look after them.
In practice you pay the learning curve once and the task or credit bill every month. If your automation footprint is going to grow, n8n’s steeper start usually pays for itself. If it’s three simple connections, don’t overcomplicate things.
Plenty of companies mix tools: simple notifications stay in Zapier while heavier processes move to n8n. That’s a sensible stepping stone, but long term two systems mean two places where things can break. If you’re new to n8n, start with what n8n is and how it works.
Migration makes sense when your task or credit bill grows faster than your number of processes, when you’re adding workaround steps to dodge platform limits, or when legal starts asking where customer data ends up. A safe plan looks like this:
For the kinds of processes we migrate most often, see 15 AI automation examples for business.
Not sure whether to stay on Zapier or Make or move to n8n? See how we deliver AI automation with n8n, or send us your list of processes and we’ll tell you where it will be cheaper and more reliable.
For multi-step, high-volume processes, n8n usually comes out cheapest because it counts one execution per full run, and you can self-host the Community edition with no licence fee. Make is cheap to start with (10,000 credits on Core), but every step uses credits. Zapier has the highest cost per action, but it is the fastest to get running.
No. Make and Zapier are cloud-only services. Of the three, only n8n can be self-hosted, both as the free Community edition and on the paid Business and Enterprise plans.
n8n gives you the most control: any model, your own API keys, Guardrails, human approval for tool calls and two-way MCP. Make AI Agents work well if your scenarios already live in Make. Zapier Agents are the simplest to set up but come with a separate activities quota (1,500 a month on Pro).
Each successful action uses one task, while triggers and polling are free. Filter, Formatter, Paths, Delay, Looping, Sub-Zap, Digest and Storage steps are free too. AI by Zapier steps use 1, 3 or 5 tasks depending on the model tier, and each Zapier MCP tool call uses 2 tasks.
It is when your task bill grows faster than your number of processes, when you keep adding workaround steps, or when customer data needs to stay on your own infrastructure. Migrate in stages: start with the Zaps that use the most tasks, then run old and new workflows side by side for one or two weeks before switching over.
Insights

n8n is an AI-ready automation platform you can self-host. Learn how it works, what it costs, what its licence allows and how to build your first workflows.
Read
15 practical AI automation examples for business: leads, invoices, reviews and customer service, with the n8n steps, the AI role and where humans decide.
Read
How much does a website cost in 2026? Indicative price ranges, what drives a quote, yearly running costs and hidden extras to ask about before you sign.
ReadWe connect your tools — CRM, email, spreadsheets, store, accounting — into n8n workflows where AI reads, classifies and replies. You get hours back, not another app to babysit.
We use essential cookies and, with your consent, analytics (Google Analytics, Hotjar) and marketing cookies to understand how you use the site and improve it. Details in our privacy policy.