Zapier vs Make vs n8n vs custom code: which should you use?
Zapier is quickest to start, Make is cheaper for multi-step flows, n8n can be self-hosted, and custom code wins at volume. Prices as of October 2026.
By Sohail Hussain21 November 20266 min read
Use Zapier when a non-developer needs to connect popular apps quickly and volume is low. Use Make when flows have several steps, branches and data reshaping, and you want to pay less per run. Use n8n when you want to self-host, keep data on your own server, or run long workflows where per-step pricing would add up. Write custom code when the flow is business-critical, high-volume or needs proper error handling, such as invoices, payments and stock. Many businesses end up with two of these, and that is fine.
We build automations in all four as part of our APIs and integrations and AI workflow automation work, and we often move clients from one to another as they grow. Prices below are as of October 2026, taken from each vendor's pricing page.
How do the four options compare?
| Zapier | Make | n8n | Custom code | |
|---|---|---|---|---|
| You pay per | Task: each successful action (triggers are free) | Credit: most module actions use one | Execution: one full workflow run | Hosting plus developer time |
| Free option | 100 tasks a month, two-step Zaps only | 1,000 credits a month | Community Edition, self-hosted | None |
| Entry paid plan | Professional, $19.99 a month billed annually for 750 tasks ($29.99 monthly) | Core, about $9 a month billed annually for 10,000 credits | Cloud Starter, €20 a month billed annually for 2,500 executions | Build cost plus a small server |
| Where it runs | Zapier's cloud | Make's cloud | n8n's cloud or your server | Your server or cloud account |
| Best for | Simple flows between popular apps | Multi-step flows with logic | Long workflows, data control, developers | Critical, high-volume or unusual flows |
| Weak at | Cost at volume, complex logic | Learning curve, debugging large scenarios | Self-hosting is your job | Up-front cost and upkeep |
Sources: Zapier pricing, Make pricing, n8n pricing. Vendors change tiers often, so check before you commit.
How do the pricing models change the bill?
The unit each tool charges for matters more than the headline price. Take one workflow: a new order arrives, then four actions run (look up the customer in the CRM, create an invoice in accounting, send a WhatsApp message, add a row to a report). It runs 3,000 times a month.
- Zapier counts the four actions: about 12,000 tasks a month. That is far beyond the 750-task entry tier, so you would be buying a much higher task tier.
- Make counts each module, trigger included: roughly 15,000 credits, a little over Core's 10,000.
- n8n counts the run once: 3,000 executions, inside the cloud Pro plan's 10,000 (€50 a month billed annually), or unlimited if you self-host.
- Custom code barely notices 3,000 runs a month. You pay for building it and keeping it working.
The more steps a workflow has, the more Zapier and Make cost relative to n8n. To see whether any of it is worth paying for, put the manual version into our automation hours-saved calculator and compare the yearly saving with the yearly platform fee.
When are Zapier or Make the right choice?
Zapier fits when the person building the automation is not a developer, the apps are mainstream, and runs are in the hundreds a month rather than the tens of thousands. It has the widest catalogue of ready-made app connections and the gentlest learning curve. Polling triggers check for new data every 15 minutes on the free plan, every 2 minutes on Professional and every minute on Team. For a lead alert or a sign-up notification, that is plenty.
Make fits when the flow has branches, loops or data that needs reshaping between steps, and someone on the team is comfortable with a more technical visual editor. At moderate volume it is usually cheaper than Zapier for the same work.
With both, the automation lives in somebody's account. Use a company login, not a personal one, and write down what each flow does. When an automation fails at 2 a.m., the person on call should not have to reverse-engineer it.
When is n8n the right choice?
n8n charges per workflow run rather than per step, which favours long workflows. It can run on your own server, so customer data does not pass through a third party's automation cloud, which matters for some clients and some contracts. It has code steps for JavaScript and Python when the visual nodes run out, and it has become a common home for AI steps; AI workflow automation: 20 tasks worth automating first lists the ones that pay back fastest. Our practical guide to n8n goes deeper.
Two cautions. First, the licence: n8n is published under the Sustainable Use License, which allows use "for your own internal business purposes or for non-commercial or personal use". Running it for your own company is fine. Building a paid product on it needs a careful read, and possibly a commercial agreement. Second, self-hosting is a job. A small server is cheap; updates, backups, monitoring and security are not free just because nobody invoices you for them.
When should you write custom code instead?
- Volume is high. Tens of thousands of multi-step runs a month cost real money on per-task platforms and very little on a small server.
- Failure is expensive. Invoices, payments and stock need retries, duplicate protection, alerts and a daily reconciliation. You can approximate these in a no-code tool, but code does it properly. Why integrations break lists what that involves.
- The rules are complicated. GST treatment by state, multi-currency rounding, or mapping that depends on five conditions is easier to read and test as code than as a forty-step visual flow.
- A system has no connector. TallyPrime on an office computer is the classic Indian example.
- It is part of a product that your clients use.
The honest counterpoint: do not start with custom code for a process you have not proven. Build it in Zapier or Make first, run it for a few months, learn where it breaks, then rebuild the version that matters. That is cheaper than building the wrong thing well. The same logic applies at a larger scale in custom software vs off-the-shelf.
How do you move off a tool when you outgrow it?
None of these platforms exports workflows in a format the others can import, so migration is a rebuild:
- Document every flow: trigger, each step, field mappings and the edge cases someone patched in later.
- Rebuild one flow at a time on the new platform, starting with the one that costs most or fails most.
- Run old and new side by side for a week, writing to a test destination, and compare the outputs.
- Repoint webhooks. Source systems send events to the old tool's URL. Change each one, and keep a list.
- Switch the old flow off only after a clean week, and keep its history for a month.
If you are choosing today, list your three most important flows with their steps and monthly runs, then price each on the three pricing pages above. The cheapest answer is often obvious within ten minutes.
Was this guide useful?