Some links in this article are affiliate links. Automated Wealth Lab may earn a commission at no extra cost to you. See our Affiliate Disclosure.
Disclosure: This article contains affiliate links to Make and n8n. If you sign up through them, Automated Wealth Lab earns a commission at no extra cost to you. It never changes what we recommend — the math below decides the pick, and we show the arithmetic so you can re-run it with your own numbers.
Every agency pricing comparison makes the same mistake: it compares plans. “Make Core is $9, n8n Starter is €20, Make wins.” That comparison is useless the moment you run client work, because neither platform bills you for a plan. Make bills you for credits consumed across every client’s scenarios. n8n bills you for workflow executions — or, self-hosted, for a server. The only unit that matters to an agency is the one you can put on a client invoice review: cost per client, per month.
So that’s the unit this article uses. We built one stated workload model of a “typical automation client,” ran it through Make’s credit slider, n8n Cloud’s execution tiers, and a self-hosted n8n server, at 1, 5, 10, and 25 clients. Every assumption is on the table. If your clients look different, swap in your numbers — the ranking logic survives.
The workload model (steal it, then adjust it)
Before any prices: what does one client actually consume? Our model client is a small business with a lead-gen site, a CRM, and weekly reporting — the bread-and-butter automation retainer. These are planning assumptions, stated so you can challenge them, not measurements:
| Client workflow | Runs/month | Make modules per run | Make credits/month | n8n executions/month |
|---|---|---|---|---|
| Lead intake → CRM → notify | 500 | 6 | 3,000 | 500 |
| CRM sync (hourly poll) | 720 | 3 | 2,160 | 720 |
| Weekly report build | 300 | 8 | 2,400 | 300 |
| Inbox/misc automations | 200 | 5 | 1,000 | 200 |
| Total per client | 1,720 | — | 8,560 | 1,720 |
The asymmetry in that table is the entire article. Make charges per module that executes inside a run — a six-step scenario costs six credits every time it fires. n8n charges once per workflow run, whether the workflow has 3 nodes or 30. The same client is an 8,560-credit customer on Make and a 1,720-execution customer on n8n. Neither number is “usage” in the abstract; they’re the same work, metered two different ways. (This is the same metering gap we broke down in the flagship n8n vs Make vs Zapier cost comparison — here we price it the way an agency actually feels it.)
Two honest warnings about the model. First, polling is the silent killer on both platforms: an hourly poll is 720 runs a month before a single real event happens, and a 5-minute poll is ~8,600. Second, iterators multiply Make credits — a scenario that loops over 20 line items pays for the loop body 20 times. Our model assumes no heavy iteration. If your client work is iterator-heavy (invoice lines, enrichment batches), Make’s per-client number climbs and n8n’s doesn’t.
The prices (checked October 2026)
Make bills in credits, priced on a volume slider, annual billing (sources at the end):
| Credits/month | Core | Pro | Teams |
|---|---|---|---|
| 10,000 | $9 | $16 | $29 |
| 40,000 | $29 | $53 | $99 |
| 80,000 | $55 | $91 | $172.90 |
| 150,000 | $99 | $153.45 | $291.56 |
| 300,000 | $182.16 | $268.54 | $510.22 |
n8n Cloud bills in executions. List prices are in euros; US visitors see dollar figures (Starter $20, Pro $50 on annual billing):
| Plan | Price (annual billing) | Executions/month | Shared projects |
|---|---|---|---|
| Starter | €20/mo | 2,500 | 1 |
| Pro | €50/mo | 10,000 | 3 |
| Pro (50K tier) | €121/mo | 50,000 | 3 |
| Community (self-hosted) | €0 license | Unmetered | n/a |
Self-hosted n8n is the Community Edition on your own server: no license fee, no execution meter. The cost is the VPS itself — realistically $5–50/month depending on load — plus your maintenance time, which is a real cost we’ll price separately below rather than pretend is zero.
The per-client math
Applying the model client (8,560 Make credits / 1,720 n8n executions per month) at four agency sizes. For Make we use the smallest slider rung that fits total credits; for n8n Cloud the smallest execution tier that fits.
| Agency size | Make (Core) total → per client | n8n Cloud total → per client | Self-hosted n8n (infra) → per client |
|---|---|---|---|
| 1 client | $9 → $9.00 | €20 → €20 | ~$10 VPS → ~$10 |
| 5 clients | $55 (80K rung) → $11.00 | €50 (10K tier) → €10 | ~$15 VPS → ~$3 |
| 10 clients | $99 (150K rung) → $9.90 | €121 (50K tier) → €12.10 | ~$25 VPS → ~$2.50 |
| 25 clients | $182.16 (300K rung) → $7.29 | €121 (50K tier) → €4.84 | ~$40 VPS → ~$1.60 |
Read the table carefully, because three non-obvious things are happening:
1. Make’s per-client cost is flat-ish, n8n Cloud’s falls off a cliff. Make’s slider is close to linear — doubling clients roughly doubles the rung price, so per-client cost hovers around $7–11 forever. n8n Cloud’s tiers are chunkier: once 25 clients fit inside the same 50K-execution tier as 10 clients, per-client cost drops by 60% without you doing anything. Tier-step pricing rewards whoever fills the tier.
2. The rung trap. Five model clients need 42,800 credits — just over the 40,000 rung. You don’t pay for 42,800; you pay for the 80,000 rung ($55 Core). Per-client cost rises from $9 to $11 when you add clients 4 and 5, then falls again. Budgeting with the headline “$9 for 10,000 credits” figure misses this entirely. Always price your book of clients against the rung above your total, not the rate card’s first row.
3. Self-hosted infra is the cheapest line in the table — and the only one that excludes your time. At 10 clients the server is ~$2.50 per client. But somebody patches that server, rotates backups, and gets paged when a client workflow dies at 2 a.m. If you value maintenance at even one hour a month across the whole book, add your hourly rate divided by client count. At a $75/hr internal rate and 10 clients, that’s +$7.50 per client — suddenly self-hosted (~$10/client all-in) ties Make Core and beats nothing. At 25 clients the same hour is +$3/client and self-hosting wins again. The honest formula: self-hosted per-client cost = VPS ÷ clients + (your maintenance hours × your rate) ÷ clients. We wrote up what that server really costs to run in the real cost of self-hosting n8n.
The costs that aren’t in the rate card
Client separation. Agencies need walls between clients. On n8n Cloud, separation runs through shared projects: 1 on Starter, 3 on Pro. A 10-client agency on Pro cannot give each client its own project — you either share projects across clients (a governance smell when a client asks “who else can see my data?”) or run multiple subscriptions. Self-hosting removes the ceiling: separate instances or one instance with disciplined naming, your call. On Make, separation is organizational (folders/teams on the Teams plan, $29 at the 10K rung and $172.90 at the 80K rung) — workable, but the Teams premium is a per-client tax that Core doesn’t charge and Pro doesn’t fully solve.
Client churn asymmetry. When a client leaves, Make credits stop being consumed the day you switch their scenarios off. Your n8n Cloud tier doesn’t shrink — you keep paying for the 50K tier until renewal. Fixed-tier platforms punish a shrinking book; metered sliders punish a growing, inefficient one. Pick your poison based on which direction your book moves.
The migration invoice. If you start on one platform and outgrow it, you rebuild every workflow by hand — neither platform exports to the other. For an agency, that rebuild is billable to nobody. This is the strongest argument for choosing the platform your 25-client self needs, not your 2-client self. (If a client is mid-migration off a dying tool, that’s a different job — see the Relay.app shutdown migration guide.)
AI steps. Both platforms now meter AI features inside the same units (Make AI module calls consume credits; n8n counts the whole agent run as one execution but cloud tiers carry their own AI-credit allowances). Agentic client work widens n8n’s per-run advantage and widens Make’s per-step exposure. If your roadmap is “AI agent does a 12-step research task per lead,” model it before you sign the retainer.
Which platform, by agency shape
Solo freelancer, 1–4 clients, simple scenarios: Make Core. $9–29/month total, zero infrastructure, and at this scale your time is worth more than the $5–15 you’d save self-hosting. Start here: Make
Agency, 5–15 clients, mixed complexity: The crossover zone. n8n Cloud Pro tiers and self-hosted n8n both land around $10/client all-in; Make lands $10–18 depending on rung and plan. Decide on separation needs and whether you enjoy owning a server. If you self-host, n8n on a mid-size VPS is the cheapest defensible setup in this article.
Agency, 15+ clients or iterator/AI-heavy work: Self-hosted n8n, and it isn’t close — per-client infra approaches $1.60–2.50 and no meter punishes your complex workflows. Budget the maintenance hour honestly and it still wins.
Non-technical team that will never touch a server: n8n Cloud at the 50K tier once you pass ~8 clients; Make Teams if your scenarios are genuinely simple and you value its app catalog (3,500+ integrations) over unit economics.
FAQ
Is Make or n8n cheaper for agencies?
It depends on the meter, not the sticker. Per model client (8,560 credits / 1,720 executions), Make Core runs $7.29–11.00 per client per month; n8n Cloud runs €4.84–20; self-hosted n8n infrastructure runs ~$1.60–10 before your maintenance time. Make wins small and simple; n8n wins as client count or workflow complexity grows.
Why does the same client cost 8,560 units on Make but 1,720 on n8n?
Different billing units. Make charges per module execution inside a scenario run; n8n charges per workflow run regardless of node count. A 6-step automation is 6 Make credits and 1 n8n execution, every time it fires.
Can I put each client in their own n8n Cloud workspace?
Not cleanly at small tiers. n8n Cloud Starter includes 1 shared project and Pro includes 3. Beyond that you’re either sharing projects across clients, buying multiple subscriptions, or self-hosting, where separation is free.
Does self-hosted n8n really cost $2.50 per client?
The infrastructure does, at 10 clients on a ~$25 VPS. The honest all-in figure adds your maintenance time: one hour a month at $75/hr internal cost adds $7.50 per client at that size. Self-hosting wins when your client count dilutes that hour, not because the hour is free.
What happens to my Make bill when a client churns?
It falls immediately — credits are consumed, not reserved. n8n Cloud tiers are fixed until renewal, so a shrinking book temporarily raises your per-client cost there. Metered billing rewards shrinkage; tier billing rewards growth into the tier.
Should I pass platform costs through to clients?
You can, but per-client metering makes it defensible: at $7–12 per client per month, most agencies fold it into the retainer as a line item (“automation platform”) rather than invoicing raw credits. What you shouldn’t do is quote a flat platform fee from a plan headline — the rung trap will eat your margin.
Methodology
What we did: Built a stated per-client workload model (four workflows, volumes and module counts published in full above), then applied October 2026 published pricing: Make’s credit-slider ladder (annual billing) and n8n Cloud’s execution tiers, plus a VPS cost range for self-hosting drawn from published hosting breakdowns. All per-client figures are arithmetic on those published prices — division shown by the table structure.
What we did NOT do: We did not run an agency, migrate client workflows, or benchmark execution speeds. The workload model is a planning model, not telemetry from live client accounts; your clients’ module counts and polling intervals will differ, and polling-heavy or iterator-heavy books will shift Make’s numbers up materially. VPS figures are mid-range estimates within the published $5–50/month band, not quotes. n8n prices are euro list prices; US visitors see dollar figures on the pricing page (Starter $20, Pro $50, annual billing), so treat converted per-client figures as approximate.
Conflict disclosure: Make and n8n are both affiliate partners. The recommendation split above follows the arithmetic — each platform wins the segment where its meter is cheaper, including segments where the winner is the one we’d earn less from at small scale.
AWL verdict
For agency work, stop comparing plans and start comparing meters. Make’s per-module credit is the friendliest way to start (a $9 first client is real) and the flattest cost curve as you grow — expect roughly $7–11 per client forever. n8n’s per-execution meter, and especially self-hosted n8n’s missing meter, is where per-client cost goes to die: €4.84 per client on Cloud’s 50K tier at 25 clients, ~$1.60 of infrastructure self-hosted. Under 5 clients, buy Make and spend your hours selling. Over 10, or the moment a client’s workflow grows loops and AI steps, move the book to n8n — the migration you’ll otherwise pay for later is the most expensive line in this whole comparison.
Sources
- Make pricing ladder by credit volume (annual billing, slider rungs 10K–300K+): https://latenode.com/blog/make-com-pricing (verified October 2026)
- Make plan prices at 10,000 credits — Core $9 / Pro $16 / Teams $29: https://dev.to/stimlau/make-pricing-in-2026-free-core-pro-teams-and-credit-costs-lcm (verified October 2026)
- Make credit mechanics (per module call; iterators multiply): https://www.usecarly.com/blog/make-com-pricing/ (verified October 2026)
- n8n Cloud pricing — Starter €20/2,500 executions, Pro €50/10,000, Pro 50K €121 (annual), shared project limits: https://www.stackscout.net/articles/cc_20261014_180524.html (verified October 2026)
- n8n pricing plans and execution model (one run = one execution regardless of steps): https://dev.to/entechsolutions/n8n-pricing-in-2026-what-each-plan-costs-and-what-you-actually-get-2amg (verified October 2026)
- Self-hosted n8n VPS cost range ($5–50/month) and Cloud crossover analysis: https://dev.to/hostingsift/what-self-hosting-n8n-really-costs-in-2026-394a (verified October 2026)
