n8n vs Zapier vs Make: what each costs at your volume

When you compare n8n vs Zapier vs Make on price, what each tool bills for matters more than the headline rate. Zapier counts tasks (each action step that succeeds), Make counts credits (each module run, trigger checks included) and n8n counts executions: "a single run of your entire workflow", however many steps it has (n8n pricing, checked 1 November 2026). Run a 4-action workflow 8,000 times a month and Zapier needs 32,000 tasks ($289 a month, billed annually) while n8n needs 8,000 executions (€50). Below: the maths at three volumes, the upkeep hours and when self-hosting pays.

One thing to know first: n8n automation is one of the services we sell. So every price here comes from the vendors' own pages, every assumption is labelled, and in our example the cheapest tool on list price turns out to be Make.

What does each tool count as one unit?

One task, one credit or one execution. Push the same workflow through all three and each run uses 4 tasks, 5 credits or 1 execution.

Here is the workflow we price throughout. Example (hypothetical): a shop's order workflow. A new order arrives by webhook. The workflow skips test orders, tidies the delivery address, looks the buyer up in the CRM, adds a row to a Google Sheet, emails a confirmation and posts a note to Slack. That's a trigger, two helper steps and four actions.

Two details matter later. Zapier says "Failed actions are not counted", so a broken step doesn't eat your allowance. And Make's flat rate has exceptions: Make Code "uses 2 credits per second of execution time", and its built-in AI apps are billed on their own rules.

What does each cost at 300, 1,800 and 8,000 runs a month?

On list prices, Make is the cheapest hosted option at all three volumes in this example, and n8n Cloud comes second. Zapier costs the most at every volume, and the gap widens as runs grow, because it bills every step of every run. Self-hosted n8n's $14.40 looks low in the price table until you add the hours.

First, the units each volume uses:

Then the smallest plan that covers each one, per month, billed annually (prices checked 1 November 2026):

How to read it:

• Billed monthly instead, the same Zapier tiers cost $58.50, $193.50 and $433.50 (Zapier pricing). n8n Starter is €24 and Pro €60, and n8n puts the annual saving at 17% (n8n pricing). Make's page says annual billing saves 15% (Make pricing).
• Make above 10,000 credits is priced with a slider. With the slider at 40,000 credits and annual billing, Make's pricing page (checked 1 November 2026) shows Core at $29 a month, which is $0.73 per 1,000 credits. At 10,000 credits Core works out to $0.90 per 1,000, so the rate per credit drops at the bigger tier. Set the slider to your own volume before you budget.
• Free plans don't fit this workflow. Zapier Free has 100 tasks a month and two-step Zaps only. Make Free has 1,000 credits a month with a 15-minute minimum interval between runs, and 300 orders already need 1,500. n8n's smallest Cloud plan is Starter; its free option is the self-hosted Community Edition.
• Euros are converted at the European Central Bank's reference rate for 30 October 2026, €1 = $1.1172 (ECB). Your card will use its own rate.
• The self-hosted column is n8n's Community Edition, which is free, on an example server: a DigitalOcean droplet with 1 vCPU and 2 GiB of memory at $12 a month, plus weekly backups at 20% of the droplet price, so $12 + $2.40 = $14.40 (DigitalOcean pricing). n8n's own sizing table, written for OEM deployments (n8n built into another product) and based on n8n Cloud, lists 320 MB to 2 GB of memory and says "n8n isn't CPU intensive" (n8n docs).

How many maintenance hours should you budget?

For one workflow like this, we would budget 1 hour a month on any hosted tool, 2 hours at the highest volume, and 3 more hours for self-hosted n8n. These are planning assumptions. We haven't measured them across these three tools, so the table shows the hours separately and you can swap in your own.

On every tool, the hours go on:

• reconnecting an app login that expired or was revoked;
• fixing a step after a connected app renames a field or changes its API;
• reading failed runs, fixing the cause and re-running them (more runs means more failures to read, which is why the 8,000-run row gets 2 hours);
• changing the workflow when the process changes, such as a new column in the sheet.

We use the same hours for the three hosted options because the work is the same kind in each. Self-hosting adds work that n8n's docs put on you: with the Community edition "you must provide and manage infrastructure" and maintenance is "your responsibility" (n8n docs). In practice that means:

• Updates. n8n's release notes show a new minor version every week from 2.39 (8 September 2026) to 2.45 (20 October), then 3.0 on 27 October, "a major release with breaking changes" with a migration tool to run first. We'd update monthly after reading the notes, and plan a separate half day for a major version.
• Backups with the encryption key. n8n creates a random key on first launch and uses it to encrypt saved credentials (n8n docs), so the key has to be backed up with the database.
• Operating system patches, disk space for saved execution data and an uptime check.

To put a price on an hour we use $40, near the top of the senior Asia band ($31 to $41 an hour) in Accelerance's rate survey of 60 development partners (published 24 November 2025). An in-house hour in New York or London costs more. There's more rate data in our guide to outsourcing software development to Pakistan.

Total per month, billed annually, hours at $40:

The hours cost more than the plan in every cell except Zapier at 1,800 and 8,000 runs. Self-hosting costs the most of the four at 300 and 1,800 runs (by $5.40 over Zapier at 1,800), and still more than Make or n8n Cloud at 8,000, because its extra hours outweigh the free licence. At 8,000 runs, Zapier's total is about 2.7 times n8n Cloud's for the same four actions. On plan price alone it's about 5 times.

When does self-hosting n8n pay for itself?

When the n8n Cloud plan you'd need costs more than the server plus the extra hours. At $40 an hour that's $14.40 + 3 × $40 = $134.40 a month. n8n Cloud has no tier between Pro's 10,000 and 50,000 executions, and the 50,000 tier is €1,450 a year (€120.83 a month, about $135). So at our hourly rate, self-hosting breaks even once you pass 10,000 executions a month, and only just: it saves less than $1 a month. If that volume needs a bigger server, it saves nothing. DigitalOcean's 4 GiB droplet is $24 a month, $28.80 with weekly backups.

Past 50,000 executions, n8n's pricing points you toward self-hosting anyway. The Business plan (€667 a month billed annually for 40,000 executions) is self-hosted only, and the Cloud option past Pro is Enterprise, priced through sales.

If your hour costs $80, the break-even rises to $14.40 + 3 × $80 = $254.40 a month, more than any Pro tier. At that rate self-hosting costs more than Cloud at every Pro volume, and the reason to do it is control of where the data sits.

One more condition: the Community Edition comes with "almost the complete feature set" (n8n docs) under n8n's Community licence, which makes it free when only you and your organization create or edit workflows (n8n licence FAQ). If your customers will build workflows inside your product, that licence doesn't cover it. This is our reading of n8n's FAQ and isn't legal advice.

What else changes the bill?

Polling, loops, AI steps, the billing interval and how many people edit. Any one of them can change the ranking above.

Polling in Make

Make's docs say "Even when a trigger finds no new data, it still uses 1 credit for checking" (Make help). A scenario that checks for new rows every 15 minutes uses 2,880 credits in a 30-day month (4 × 24 × 30) before a single order arrives. Checking every minute uses 43,200. Zapier doesn't count polling, and n8n doesn't count polls that return nothing, though its Schedule Trigger counts every time it fires. In Make, use an instant (webhook) trigger where the app offers one.

Loops

Example: each order has 3 line items and each becomes its own sheet row. In Zapier, "each action step after the loop uses 1 task for each loop that runs" (Zapier help, updated 29 May 2026), and the Looping step itself is free. Put the loop last, so only the sheet step repeats, and an order goes from 4 tasks to 6. In Make, "each bundle triggers its own module run" (Make help). With the loop last again, the row step runs 3 times, and the Iterator module that splits the order into items uses a credit too (Make's pricing page lists "iterate a row of data" among the actions that do), so an order goes from 5 credits to 8. n8n still counts 1 execution. At 8,000 orders, that's 48,000 Zapier tasks, close to the top of the 50,000 tier, and 64,000 Make credits.

AI steps

Zapier's pricing page says "Many actions cost more than one task", and the amount "depends on step type, AI model tier, and number of tool calls". In Make, AI steps through Make's own AI Provider cost 1 credit per operation plus credits for the tokens used. Connect your own AI provider instead (paid plans only) and it's a flat 1 credit per operation, with the provider billing you for tokens (Make help). n8n lists AI nodes on every plan; with your own API key, the model provider bills you per token. Our token maths is in what an AI chatbot costs to build and run.

Monthly billing and overage

Billed monthly, every Zapier tier above costs 50% more. With pay-per-task switched on, Zapier's pricing page says extra tasks cost 2.5 times the base task rate on monthly plans and 1.25 times on annual ones, and usage stops at 3 times your plan's tasks. Its help article (updated 29 May 2026) gives a single 1.25 times rate, so check your own billing page. Task allowances reset monthly, even on annual plans.

People

Zapier Pro is for 1 user. Team allows up to 25 and costs $169 a month at 10,000 tasks billed annually, against Pro's $129. n8n says "All plans include unlimited users & workflows". Make's team roles start on its Teams plan, $29 a month for 10,000 credits billed annually.

Limits besides price

n8n Starter runs 5 executions at once and caps each at 5 minutes. Pro allows 20 or 50 at once, depending on tier, and 40 minutes. Check your busiest minute and your slowest step (an AI call on a long document, say) against those numbers.

n8n vs Make vs Zapier: when is each the wrong choice?

Each one has a case where its pricing or its design works against you.

• Zapier is the wrong choice for long workflows at volume, or for anything that loops over line items, because every step of every run is billed. It's a fair choice when an app you depend on only has a good Zapier connector, or your team already runs everything there.
• Make is the wrong choice when the trigger has to poll often and the app offers no webhook, because the checks alone can use more credits than the work. Heavy custom code is the other case, at 2 credits per second.
• n8n Cloud asks more of whoever builds and fixes the workflows. Its expressions are "small pieces of JavaScript-like code" (n8n docs), and you'll read the JSON data each step passes on. If nobody on your team is comfortable with that and you won't hire help, choose Make or Zapier.
• Self-hosted n8n is the wrong choice when nobody owns updates and backups. A server left on an old version, with the breaking 3.0 upgrade still waiting, is a cost you haven't counted yet.
• None of them fits a process that still changes every week. Automating it freezes it in place, so run it by hand until it settles.

Before you pick, check each tool's app directory for the exact trigger and action you need, such as "new paid order" or "update deal stage". An app's name appearing in the directory doesn't mean the step you need exists.

What we would do

This is how we'd decide, with the conditions that would change it:

1. Under about 2,000 runs a month (where Make Core's 10,000 credits run out for a 5-credit workflow), short workflows, every app supported: start on Make or n8n Cloud Starter. In our example the difference is $13.34 a month, so choose the one whose connectors cover your apps and whose editor your team finds easier.
2. Workflows with many steps, loops or AI agents, or several people editing: n8n Cloud. Billing per run keeps the bill flat as steps get added, and every plan includes unlimited users.
3. Past 50,000 executions a month, where n8n Cloud's next step is Enterprise, or data that must stay on your own servers: self-host n8n, with a named person for updates, backups and the encryption key. Between 10,000 and 50,000, Cloud and self-hosting cost about the same at $40 an hour, so choose on who will look after the server.
4. Already on Zapier: work out the payback before you move. Every workflow gets rebuilt and tested again. Example: if a rebuild takes 10 hours at $40, that's $400, which takes 4 months to recover at $100 a month in savings and more than a year at $30.

If you'd like someone to do the counting, the build and the upkeep, our n8n workflow automation service starts with a scope call and a fixed written plan that lists the running costs separately.

How to price your own workflows

Price your own workflows the same way, one at a time:

1. Write down each workflow: its trigger, every step and how often it runs in a normal month and a busy one.
2. Count units per run: Zapier action steps, Make modules (trigger included, and every polling check), and 1 execution for n8n.
3. Multiply by runs and add the extra runs from loops over line items.
4. Pick the smallest tier that covers your busy month on each vendor's page, on the billing interval you'll really use.
5. Add your maintenance hours at your own hourly cost, plus 3 hours if you'll self-host.
6. Write the date you checked the prices next to every figure. All three vendors list several tiers, and prices change.

Then apply one rule: if the cheapest option saves you less each month than one hour of your own time, choose the tool your team can fix without outside help.