n8n vs Make: what a 10-step workflow costs on each

In n8n vs Make, the price gap comes from what each one counts. Make bills credits, and most module actions cost 1 credit, so a 10-step workflow run 10,000 times uses about 100,000 credits a month. n8n bills executions, one per full run, so the same job is 10,000 executions: exactly n8n Cloud Pro (50 € a month, billed annually), or free software on a 6.64 € server that ran our test workflow at 15.04 executions/s.

This matters now because Make switched its billing unit from operations to credits on 27 August 2025 and n8n 2.x reshaped its plans, so many comparisons still use old terms. We read both official pricing pages on 5 November 2026.

Tested in the Lab Hetzner CX23, CX33, CX43, CAX11, CAX21 · Nov 2026

This page sits in our n8n hosting hub. We have run n8n ourselves; we have not measured Make, since it cannot be self-hosted, so everything about Make here comes from Make's own pages.

The two billing models, from the official pages

Make's help centre adds two details. Existing operations were converted 1:1 into credits, and on annual Pro and Teams plans you get the year's credits as one pool instead of a monthly allotment; Core always gets monthly credits.

One workflow, priced both ways

Take a typical 10-step workflow: a webhook trigger, a lookup in your CRM, a filter, a format step, an AI summary through an OpenAI module, a database insert, a Slack message, an email, a log row and a response. Assume every step is a module that processes data, and the workflow runs 10,000 times a month.

Make's per-step model means the same job costs ten times the units it would at one step per run. If you add a router or error handler, those do not count; if you use Make's built-in AI Provider instead of the OpenAI module, the AI step can cost more than 1 credit.

About Make's price for that tier: on 5 November 2026 its pricing page set prices through an interactive credit selector that our tools could not read, so we do not quote a figure here rather than guess. The decision rule is simple. Open the selector, choose the tier that covers your credits, and compare it with 50 € for n8n Pro, or about 8 € plus your time for self-hosted n8n. Annual billing is cheaper on both; n8n states a 17% saving.

Now reverse the example. A 2-step scenario run 400 times a month uses about 800 credits, which fits Make's free 1,000. On n8n that is 400 executions: no free Cloud plan, so either Starter at 20 € a month or a server you look after. Short, low-volume automations are where Make is the fair winner.

Self-hosting n8n: how much a small server handles

The self-hosted column only makes sense if a cheap server keeps up. We measured that. n8n 2.46.6 with Postgres 16 ran a webhook → Code node → response workflow under ApacheBench on five Hetzner servers:

Measured by AI Hosting Lab on 2026-11-05 · 5 servers · method · raw data

The 2 vCPU, 4 GB CX23 peaked at 15.04 executions/s with zero failed requests. At that rate, 10,000 runs of our test workflow add up to about 665 seconds of work in a month. Your 10-step workflow will be slower per run, because each external call waits on the network and on the other service, but even at fifty times our per-run time, that is about 9 hours of work spread over a month. The 8 GB CX33 reached 31.93 executions/s if you want more headroom. Prices, from the Hetzner Cloud API:

From the Hetzner Cloud API, location hel1, checked 2026-11-05. "Incl. VAT" is what our Czech (EU consumer) account pays; businesses and buyers outside the EU usually pay the price excl. VAT. Primary IPv4 is billed separately.

Want the full cost picture, backups and your time included? Use the calculator: enter the Cloud plan or Make tier you are comparing as the "Cloud plan price", your server price and your upkeep hours.

More on the cost lines in n8n hosting price, and the Cloud vs own-server trade-offs in n8n Cloud vs self hosted.

Make: the fair case

• Best for: non-developers and small teams who want a polished visual builder, short scenarios and no servers at all.
• Main drawback: every module run costs a credit, so long workflows and loops over many items get expensive; scenarios stop when credits run out until you add more.
• Skip it if: your workflows have many steps, iterate over large lists, or must keep data on your own infrastructure.

Make says it has over 3,000 pre-built app integrations (its own figure). You get no server to patch, no backups to check and no TLS certificates to renew. Usage allowances scale with credits: per 10,000 monthly credits Make lists 5 GB of data transfer, 10 MB of data storage and a webhook queue of 667. Extra credits come in bundles of 1,000 or 10,000 and expire after one month on monthly plans. See Make pricing (no commission for us).

n8n Cloud: the middle option

• Best for: teams with long workflows who do not want a server.
• Main drawback: fixed execution allowances (2.5K on Starter, 10K on Pro) and capped concurrency (5 on Starter).
• Skip it if: your volume is high enough that a 6.64 € server plus an hour of your time is far below the plan price.

Plans: Starter 20 €, Pro 50 €, both billed annually, unlimited users and workflows. n8n Cloud, again no commission.

n8n self-hosted: the volume option

• Best for: high volume, long workflows, AI agents and anyone who wants data on their own machine.
• Main drawback: you own updates, backups and security. The free Community Edition also lacks projects, sharing, SSO, environments and Git version control.
• Skip it if: nobody on your team will open a terminal once a month.

Our production n8n on a CX23 used 1036 MiB of system RAM idle (MiB: mebibyte, 1,048,576 bytes) and installed in 127 seconds. Server: Hetzner Cloud, no commission.

Beyond price

A few differences that matter once you build real things:

• Code. n8n has a Code node for JavaScript inside a workflow; our load test runs one through n8n's task runner. Heavy data shaping costs no extra units there.
• Loops. On Make, iterating over 100 items through 3 modules is about 300 credits. On n8n it stays one execution.
• Data location. Make runs only on Make's cloud. Self-hosted n8n keeps payloads on your server.
• Other tools. Zapier, Activepieces, Windmill and Node-RED sit in the same space; we cover them separately as we test them.

Which one for which job

How to compare with your own workflow

1. Count the modules in your busiest scenario that process data. Leave out routers and error handlers, which Make does not charge.
2. Multiply by the runs per month. If a module iterates over items, multiply by the items too, since each module run uses a credit.
3. Find the Make tier that covers that credit total on Make's pricing selector.
4. Count your n8n executions: one per run. Compare with Starter (2.5K) and Pro (10K).
5. For self-hosting, enter the Cloud price you would otherwise pay, your server price and your hours into the calculator above.

Two checks people skip. On Make, a scenario that uses Make's own AI Provider can use more than one credit per AI step, so test a run and read the credit count in the scenario history. On n8n Cloud, the Starter plan runs at most 5 executions at once and queues the rest, which matters if your trigger is a busy webhook. Our self-hosted test ran 50 concurrent requests with zero failed requests on every server.

Next step: if n8n came out ahead for your workflows, pick a server in our best VPS for n8n guide, ranked by the same load test shown above.