n8n and Make bill the same work in different units, and that one difference decides which is cheaper for a small agency. n8n charges per execution: one whole run of a workflow, however many steps it contains. Make charges per credit: one module action inside a run. On the US list prices we checked on November 11, 2026, Make's Core plan costs $16 a month for 10,000 credits and n8n's Starter costs $24 for 2,500 executions. Make is cheaper while workflows stay short; n8n wins once a workflow passes roughly four to seven module actions per run, and it wins outright if you want to host client data yourself.

How we researched this: Documentation-based research, not a hands-on test. We read n8n's pricing page and its execution documentation, and Make's pricing page and credits documentation, on November 11, 2026. Prices are US dollars as displayed to a US-priced session; both vendors localize prices or currency by country, so open the pages from your own market. Neither platform was connected to a client account. See how Hostdebitmedia evaluates software.

The billing unit is the whole decision

A Make credit is consumed by a module action: reading a spreadsheet row, creating a record, sending an email. Make's pricing FAQ says router modules and the error-handler modules (rollback, break, resume, commit, ignore) do not consume credits, and that AI modules can consume credits based on token use as well as operations.

An n8n execution is a complete run. n8n states that pricing is based on monthly workflow executions "regardless of complexity" and that a workflow with unlimited steps still costs one execution per run. Its documentation lists what does not count against the quota: manual editor runs, sub-workflow runs (only the parent counts), error-workflow runs, polls that find no new data, and webhook requests that fail before the workflow starts.

Make n8n
Unit One credit per module action One execution per run
What one unit covers A single step in the scenario Every step from trigger to finish
Not billed Router modules, the five error-handler modules Manual runs, sub-workflow runs, error-workflow runs, empty polls, malformed webhooks
Still billed Every module action performed, including repeats inside a loop A scheduled run even if it fails; every webhook request that starts the workflow, including an empty body
AI steps Credits vary with tokens as well as operations Covered by the run; Assistant credits are a separate allowance

The practical consequence is that a ten-step Make scenario costs about three times the credits of a three-step scenario, while the same two n8n workflows cost the same per run. Loops behave the same way: a Make iterator that runs one module per client consumes a credit per client, while the n8n loop stays inside a single execution. That is the entire cost model in one sentence.

Where the two bills cross

The entry plans make the comparison concrete. Make Core includes 10,000 credits; n8n Starter includes 2,500 executions. Divide each price by its allowance to get a cost per unit, then multiply the Make figure by the number of module actions a run performs.

Plan (monthly billing) Price Included units Cost per unit 3-step run 10-step run 25-step run
Make Core $16 10,000 credits $0.0016 $0.0048 $0.0160 $0.0400
n8n Starter $24 2,500 executions $0.0096 $0.0096 $0.0096 $0.0096
n8n Pro $60 10,000 executions $0.0060 $0.0060 $0.0060 $0.0060

Read the last three columns as marginal cost, not as a bill: at low volume the plan fee dominates and neither figure is what you pay.

The break-even is the number of module actions at which a Make run costs the same as an n8n execution. Against Starter, that is 0.0096 divided by 0.0016, or six steps per run. Against Pro it is 3.75 steps. With annual billing, the same arithmetic gives about five and four steps. In practice: below four to seven module actions per run, Make is cheaper; above that, n8n is.

One more number matters before the step count: the price of the next tier. On n8n, moving from 2,500 to 10,000 executions means moving from Starter to Pro, which adds $36 a month at monthly billing. On Make, moving from 10,000 to 20,000 credits adds $12. If you expect steady growth, Make's tiers are gentler; if you expect each workflow to grow longer rather than more frequent, n8n's single tier absorbs it.

1,200 runs a month: where the answer flips

Agency workflows are not all the same length, so the honest test is a fixed number of runs with a changing step count. At 1,200 production runs a month, about 40 a day, n8n stays on its $24 entry plan: 1,200 executions is under half of Starter's 2,500. Only Make's plan moves.

Steps per run Make credits used Make plan credits Make price n8n price Cheaper
3 3,600 10,000 $16 $24 Make
8 9,600 10,000 $16 $24 Make
12 14,400 20,000 $28 $24 n8n
25 30,000 40,000 $51 $24 n8n

Two things stand out. First, the flip happens between eight and twelve steps per run, which is exactly the range of a realistic onboarding or reporting workflow. Second, the eight-step row uses 96% of Make's 10,000 credits: one retry loop, one extra client, or one month of testing pushes it into the next tier. Make's price for that next tier is $28 at 20,000 credits and $51 at 40,000, read from the pricing data behind the selector on the same date.

The same three workflows on both platforms

This is the model we used in the Make pricing explainer, applied to both platforms. It is illustrative, not measured usage, and it assumes one credit per non-AI module action with no failed runs.

Workflow Modules per run Runs a month at 10 clients Make credits n8n executions
New-client onboarding 8 10 80 10
Approval reminders 5 40 200 40
Monthly client reporting 9 10 90 10
Total 60 370 60

At five clients the same model produces 185 credits and 30 runs. At ten clients, 370 credits and 60 runs: 3.7% of Make's 10,000-credit allowance and 2.4% of n8n's 2,500 executions. The marginal cost of the units used is almost identical, $0.59 on Make and $0.58 on n8n, because the average run is six module actions long. What differs is the plan you must buy: $16 against $24.

So for a ten-client agency running three simple workflows, the decision is not about volume. Both entry plans have more than 25 times the headroom required. It is about the other gates below, and about what happens when a workflow grows.

Plan gates that change the answer

Gate Make n8n
Active workflows 2 on Free, unlimited on paid plans Unlimited on every plan
Maximum run duration 5 minutes on Free, 40 minutes on paid plans 5 minutes on Starter, 40 minutes on Pro and Enterprise
Concurrency No concurrency number published 5 concurrent on Starter, up to 50 on Pro, 200+ on Enterprise
Execution log retention 7 days on Free, 30 days on paid plans 7 days on Starter, 30 days on Pro, unlimited on Enterprise
Saved runs Not published as a run count 2,500 on Starter, 25,000 on Pro, 50,000 on Enterprise
Self-hosting Not offered; Make hosts on AWS in the EU or North America Free Community Edition; the Business plan is self-hosted only
Git version control Not listed in Make's plan comparison Listed from Business and on Enterprise
SSO and directory sync Enterprise only Enterprise on Cloud; the self-hosted Business plan lists SSO and LDAP
Overage Buy credit bundles or auto-purchase; overage protection is Enterprise-only Capped on Starter and Pro, custom on Enterprise

Error handling is priced differently too. Make's error-handler modules are free, so a retry path does not add credits, while n8n includes automatic retries on every plan and does not count error-workflow runs against the quota. Both designs reward putting failure handling into the workflow rather than leaving it to a person watching a screen.

The 5-minute ceiling on n8n Starter matters more than the price. A reporting workflow that fetches five analytics APIs, waits on each and then assembles a document can exceed five minutes, and n8n's own pricing page puts the 40-minute limit behind Pro. Make gives paid plans 40 minutes from Core upward, so a slow workflow is a price problem on Make and a plan-upgrade problem on n8n.

Data, ownership, and client trust

Make is cloud-only. Its pricing page states AWS hosting in the EU or North America, and SSO and audit logs are Enterprise features. That is a normal SaaS arrangement: you accept the vendor's infrastructure, and you get their uptime work in return.

n8n publishes a self-hosted Community Edition on GitHub, and its Business plan is self-hosted only. For an agency handling client credentials, that is the strongest structural difference in this comparison: the data stays on infrastructure you control, and you can put workflows under version control. The trade is explicit — you own patching, backups, uptime and the person who gets paged when the server stops. A two-person agency without an operations owner should price that work before choosing the self-hosted route.

Neither platform's entry plan includes SSO, so agencies with a security questionnaire asking for single sign-on are negotiating Enterprise on both sides.

A two-week trial you can run

This is a proposed procedure, not a test we performed. Run the same workflow on both platforms with fictional client data. Owner: the person who will maintain automation after the trial.

# Test Why it decides Pass condition
1 Run one onboarding workflow 20 times Shows the real unit consumption Make records credits per run; n8n records one execution per run
2 Break a step deliberately Shows how failures are billed and how visible they are The failure is visible in the log the same day
3 Run the workflow for ten clients Tests scaling without copy-paste One template, ten client configurations
4 Have a non-builder change one field Tests handoff The change is made without developer help
5 Pull the log for one client question Tests the audit trail Step-level detail exists inside the retention window
6 Price your busiest month Tests plan headroom The plan covers twice your forecast without a tier change

Record four numbers at the end: unit consumption per run, setup hours, monthly cost at your real volume, and the cost of your worst month. Stop rule: if either platform needs a plan change inside the first two weeks at real volume, recompute before signing an annual commitment. Annual billing cuts Make's 10,000-credit price from $16 to $12 a month and n8n's Starter from $24 to $20, which is not enough to justify a year on the wrong billing unit.

When to choose Make

  • Workflows are short and stable: three to six module actions, run on a schedule.
  • You want a visual builder that a non-technical account manager can edit.
  • You need a permanent free tier: 1,000 credits and two active scenarios.
  • You want module-level pricing you can predict from a diagram.
  • You are comfortable with vendor-hosted infrastructure.

When to choose n8n

  • Workflows are long, branched, or code-heavy, so per-step billing punishes you.
  • You test constantly; manual runs do not consume the quota.
  • Client data must stay on your own infrastructure.
  • You want Git version control and multiple environments.
  • You expect the automation to become a service you sell, not a back-office helper.

Frequently asked questions

Is n8n or Make cheaper for a ten-client agency?

Neither, at that size. Our illustrative three-workflow model uses 370 Make credits or 60 n8n executions a month, far inside the entry plans. The bills are $16 for Make Core and $24 for n8n Starter on monthly billing. The choice should be made on workflow length, hosting and handoff, not on the entry price.

Does Make charge per step?

Yes, in effect. One module action consumes one credit, and routers and error handlers are the documented exceptions. A ten-step scenario therefore costs about ten credits per run, while a three-step scenario costs three. Make's credit selector sets the price at each volume; at 10,000 credits the US list price is $16 a month.

Does n8n charge for failed runs?

It depends on the trigger. n8n's documentation says a scheduled run counts every time the node fires, regardless of outcome, so a failing scheduled workflow still consumes one execution. Polling triggers count only when new data is found, and webhook requests that fail before the workflow starts do not count.

Can n8n run on my own server?

Yes. n8n publishes a free self-hosted Community Edition on GitHub, and its Business plan is self-hosted only. Make does not offer self-hosting; it runs on AWS in the EU or North America. Self-hosting moves patching, backups and uptime to your team, so price that work against the subscription saving.

Do both platforms include all integrations?

n8n states that every plan includes every integration, with unlimited users and workflows. Make lists 3,500+ standard apps plus custom apps on every plan, with enterprise apps reserved for Enterprise. Neither platform charges per seat on the plans compared here.

Which is easier for a non-technical team?

That is a judgment, not a documented fact. Make's published strength is a visual scenario canvas built from app modules. n8n's documentation emphasises code steps, custom nodes and Git-based version control, which assume a technical owner. Test it directly with test 4 above: ask the person who will maintain the automation to make one change.

Start with the client onboarding hub for the process these workflows support, then read the Make pricing explainer if Make's credit model is the one you are pricing. The Make onboarding automation guide lists the modules that consume credits, and the automation audit checklist helps you decide which jobs are worth automating at all. If you are not buying a platform yet, the client onboarding template covers the same process manually, and Make vs Zapier covers the other common comparison. Browse all software guides for the rest.

Next step: open both pricing pages, set each to your own country's currency, and write down the price of the plan that covers twice your current volume. Then run test 1 for one workflow on each platform. If the smaller number is on the platform with the longer workflow, the billing unit is doing the work for you.