home/insights/zapier-make-n8n-or-custom-code

Automation16 Nov 20268 min read

Zapier vs. Make vs. n8n vs. Custom Code: The Honest Comparison

Zapier runs on tasks, Make hides its prices behind a sales call, and n8n isn't quite open source. Here's what each one actually costs and when to skip all four.

Markus BehmannFounder & Chief Unicorn Officer

In plain English: Zapier, Make and n8n price themselves three different ways (tasks, 'operations', and 'executions'), which makes them almost impossible to compare on a spec sheet. Pick by what breaks first for you — integration count, execution volume, or control over the logic — not by whichever page has the shiniest free tier.

In March 2025, n8n — one of the four tools in this post's title — raised $60 million. Seven months later, in October 2025, it raised another $180 million at a $2.5 billion valuation, with Nvidia's venture arm among the backers (Highland Europe; Sifted). That's not a company selling a nice-to-have. That's investors betting that "which tool wires your systems together" is a multi-billion-euro question.

And yet, if you try to actually compare the four obvious answers — Zapier, Make, n8n, or just paying someone to write the integration — you'll spend an afternoon and end up less sure than when you started. Not because the information doesn't exist. Because each vendor prices itself in a different unit, on purpose, so you can't put the numbers side by side without doing the conversion yourself. This post does that conversion.

The three pricing languages, translated

Here's the trick first, because it explains everything else: Zapier charges per task (one action in one automation, triggered once), Make charges per operation, now rebranded "credit" (roughly one step in one automation run), and n8n charges per workflow execution (one full run of a workflow, however many steps it has). These are not the same unit, and no vendor translates one into another for you.

ZapierMaken8n (Cloud)
Pricing unitTasks/monthOperations ("credits")/monthWorkflow executions/month
Free tier100 tasks, 2-step zaps only1,000 credits1,000 executions (trial)
Entry paid tier750 tasks ≈ $19.99/mo (annual)Unpriced on the public page — see below2,500 executions, €20/mo
Top listed cloud tier2,000,000 tasks ≈ $3,389/mo (annual)Unpriced on the public page40,000 executions, €667/mo
Self-hosting optionNoNoYes, free (Community Edition)
App/integration count claimed10,512+3,500+~2,700 listed

Sources: Zapier pricing, Zapier apps, Make pricing, Make integrations, n8n pricing, n8n integrations.

Notice the gap in the Make row. That's not an oversight on our part — Make's own pricing page names the Core, Pro, Teams and Enterprise tiers and lists their credit allowances, but doesn't show a single dollar figure without you first picking a region and a seat count the page loads separately. For a company whose entire pitch is "no-code, see the whole thing on one screen", that's a small, honest bit of irony. If you want a real number, you'll get it from a sales conversation, not a page load — so budget that call before you budget the tool.

What "cheap" actually means per tool

A free tier tells you almost nothing, because the three platforms hit their ceiling at wildly different points:

  • Zapier's free tier stops you at two-step automations entirely. You can't chain three actions together without paying, no matter how few times a month you run it.
  • Make's free tier gives you 1,000 operations a month, but multi-step scenarios burn through them fast — a five-step scenario running twice a day is already 300 operations a month before you've automated anything that actually matters.
  • n8n's free option isn't really the cloud trial (1,000 executions, then you're done) — it's self-hosting the Community Edition, which has no execution cap at all. The catch: you now own a server.

That last point matters more than the marketing suggests. "Free and open source" doesn't mean free. A small VPS to run n8n yourself starts at roughly €5.49 a month excluding VAT for the cheapest shared-resource plan from a mainstream European host (Hetzner Cloud) — plus the hours someone on your team spends keeping it patched, backed up and awake at 3am when it isn't. That's a real cost. It's just one that doesn't show up on n8n's pricing page, because it isn't n8n's to show.

There's a second wrinkle worth knowing before you build your whole stack on it: n8n describes itself as "fair-code", distributed under its own Sustainable Use License rather than a standard open-source licence (n8n's GitHub README). In plain terms, you can read it, self-host it and modify it freely — you just can't resell it or offer it as a competing hosted service without a commercial agreement. For 99% of SMEs using it to run their own workflows, that distinction changes nothing. It's worth knowing before you tell your board it's "open source" and someone who reads licences corrects you in the meeting.

// quick one

Want us to do this for you? We only take a few new clients at a time.

See if we’re a fit

Where each one actually breaks

Every one of these tools is genuinely good at what it's built for. The honest question isn't "which is best" — it's "which one breaks first for your situation, and what does that cost you." Three failure modes show up in practice:

Integration coverage. If the two systems you need to connect both have well-behaved, popular APIs, Zapier's 10,500+ app library (the widest of the three, per its own count) will almost certainly have both already built, tested and maintained by someone else. That maintenance is worth more than it sounds — Zapier's own engineering team has written publicly about how much ongoing effort it takes to keep direct API integrations working as providers quietly change or retire their endpoints, concluding plainly that "building a bunch of direct integrations is usually easier than maintaining all of those integrations" (Zapier Engineering). That's a vendor's argument for its own product, but it's also just true: somebody has to notice when an API breaks, and it's nicer when that somebody isn't you.

Logic complexity. No-code tools are built around linear or lightly-branching steps: if this, then that, maybe with a filter. The moment your process needs real conditional logic, loops that depend on each other's output, or state that persists across runs, you're fighting the tool's canvas instead of using it. David Hsu, founder and CEO of the low-code platform Retool, put a number on this himself in a 2022 TechCrunch roundtable of named industry figures: no-code and low-code tools can typically cover "50%, 60%, or 70%" of a business's workflow needs without writing code, leaving the rest — the parts specific to how your business actually works — needing custom logic (TechCrunch, March 2022). Patrick Jean, CTO of OutSystems, made the same point from the other direction: these tools "often run into roadblocks when they need to scale or evolve over time." Worth noting: both are executives at companies that sell adjacent tools, so take the framing, not the precision, as the useful bit.

Volume economics. Task- and operation-based pricing is fine until it isn't. Run the maths on Zapier's own published tiers: 750 tasks a month costs roughly $20; scale that same linear logic up and the highest listed tier, 2,000,000 tasks, costs around $3,389 a month on annual billing. Somewhere between those two numbers, for a specific business, the monthly tool bill crosses what it would cost to have someone build and host the equivalent as plain code. Nobody — not Zapier, not Make, not n8n, and not the agencies that sell custom integrations either — publishes a clean, general "crossover point" for when that happens, because it genuinely depends on how complex your logic is and what your team's time is worth. Treat any blog post that gives you a precise number here (including, if you ever write one, your own) with real scepticism.

A quick decision rule

You don't need a consultant to pick between these four. Ask, in order:

  1. Does a pre-built connector already exist for both systems, and is the logic basically linear? Use Zapier or Make. Pick whichever one's free tier survives your actual step count — check the table above, not the marketing page.
  2. Is the volume high, or do you want the workflow logic itself visible and editable by more than one person, without per-seat pricing? Look at n8n, and budget for either its paid cloud tier or a small self-hosted server plus someone responsible for it.
  3. Does the logic branch in ways none of the above canvases handle cleanly, does it need to talk to your own database directly, or does it need to be fast and auditable at a scale where per-task billing becomes real money? That's custom code — not because no-code tools are bad, but because you've found the 30–40% Hsu was talking about.
  4. Are you not sure which of the three you're in? That's usually a sign you haven't actually mapped the workflow yet. Our two-hour exercise for scoring and picking your first automations is built for exactly that gap, before you've committed to a tool at all.

The part nobody publishes

We looked for a solid, published benchmark on what a custom integration project actually costs in Europe, specifically, so this section could end with a number instead of a shrug. It doesn't exist. General freelance developer day-rate data is out there — a UK freelance marketplace reports an average day rate around £509 across IT roles generally (Malt) — but that's a rate for a person's time, not a price for an integration project, and treating the two as interchangeable would be exactly the kind of invented precision we're warning you against above.

That gap is itself useful information. Nobody who's been burned by an overrun automation project writes the honest case study afterwards, and nobody selling a fixed quote wants to publish the number that lets a prospect negotiate against it before the call even starts. If a comparison like this one feels incomplete at exactly the point where you need a hard number, that's not you missing something — that's the industry.

It's also where buying, bending or building stops being a tooling question and starts being a company-specific one: what does the hour actually cost you if nobody does this work, versus what a fixed price looks like once someone's actually seen your systems. We build this kind of thing as part of our AI & process automation work, and the custom-code end of it as custom software — in both cases, the first conversation is free, and it starts with mapping what you've actually got, not with picking a tool off this table.


Not sure yet whether your bottleneck is tooling or something further upstream? The scorecard takes ten minutes and tells you which one.

Written by Markus

Founder & Chief Unicorn Officer at VEONIO. Leads every engagement personally. Still can’t say no to a good idea.

// Q4 2026 intake · 3 spots left

Ready for steroids*?

Apply in 2 minutes. If we’re a fit, you get a free strategy session with three concrete wins for your business. If not, we’ll tell you who is.

Apply to become a client If you love it, hit this button* · 2 min, free strategy session

* 100% legal. Pinky promise.