n8n vs Make: Self-Host Unmetered vs Flat-Fee Cloud

Make charges per credit and runs itself. n8n charges per execution on Cloud — or nothing per run if you self-host a $5 VPS and absorb the ops. The right answer flips hard with volume.

The core difference: per-credit vs per-execution vs unmetered

These two tools price work in fundamentally different units, and that single fact decides almost everything downstream.

• Make bills per credit. Every module (step) that processes data usually consumes at least one credit. A workflow with eight steps burns at least eight credits per run (source: make.com/en/compare/make-vs-n8n, 2026-05-30).
• n8n bills per execution — one whole workflow run counts as one, no matter how many nodes fire (source: make.com comparison, 2026-05-30).
• n8n self-hosted doesn't bill per run at all. The software is open-source (fair-code, Sustainable Use License) and self-hostable with unlimited executions — you pay only for the server (source: github.com/n8n-io/n8n, 2026-05-30).

So the comparison isn't really "n8n vs Make." It's three cost models: Make's metered-but-zero-ops cloud, n8n's metered cloud, and n8n's flat-fee self-host. The metering unit is why the cheap option flips as you scale.

Cost at 5k / 50k / 500k runs per month

This is the wedge. Same workload, three volumes, real published prices where they exist.

Sources: make.com/en/pricing, n8n.io/pricing, github.com/n8n-io/n8n, all as of 2026-05-30.

Read the rows carefully:

• At 5k, Make ($9) is cheapest among the no-ops options. n8n self-host is technically cheaper still (~$5) but only if a VPS is worth your time at this size — usually it isn't.
• At 50k, n8n Cloud is the worst option on the board. Business includes only 40k executions (EUR 667/mo), and overage is sold in 300k buckets at EUR 4,000 each — so 10k of spillover means either buying a full EUR 4,000 bucket or moving to Enterprise. Make stays modest. n8n self-host barely moves.
• At 500k, self-host is an order of magnitude cheaper than either cloud — if you can run it.

Honesty notes: the Make 50k figure isn't flat-published — the slider jumps 40k to 80k and the exact sticker only appears live. The Make 500k ~$338 figure comes from a secondary source (dev.to), not confirmed on Make's live slider as of 2026-05-30. n8n Cloud's exact 50k overage outcome depends on negotiation, and the 500k Enterprise price is not publicly listed.

Ease and ops: the number behind the cheap VPS bill

A $5 server bill is only the visible part of self-hosting. At low volume, a single n8n container on a Hetzner CX22 is genuinely light. At scale you move to queue mode, and queue mode means running Postgres + Redis + a main process + worker processes — and keeping all of them healthy (source: github.com/n8n-io/n8n/issues/17395, 2026-05-30).

That's not theoretical. Documented production failure modes include stuck executions and Redis idle-timeout drops. Two real reports from the n8n repo:

ALL workflows fail to execute when n8n is running in queue mode (N8N_EXECUTIONS_MODE=queue). The error occurs immediately when any workflow is triggered (webhooks, manual execution, scheduled triggers, etc.), preventing the execution from being queued and processed by worker containers.

hybridmindset, GitHub issue #17395, 2025-07-16

Add Redis Keep-Alive Configuration for Queue Mode

LostInBrittany, GitHub issue #21319, 2025-10-29 — feature request: server-side idle timeouts (~10 min) terminate Redis connections in queue mode, disrupting job processing

Make has none of this. It's a hosted cloud — zero servers, zero Redis, zero key rotation. That zero-ops property is the thing you're actually buying when you pay Make's per-credit premium at high volume. Estimate ~1-2 maintenance hrs/mo for a small self-hosted instance and ~2-3 hrs/mo once you're in queue mode (these are planning figures, not a cited source).

Integration depth and where each is stronger

Make states 3,000+ app integrations. n8n states 1,200+ pre-built apps (source: make.com comparison, 2026-05-30).

On raw prebuilt-app count, Make is broader. But the count understates n8n for technical teams: n8n leans on code nodes and raw HTTP, so anything with an API is reachable even without a dedicated connector. Practical read: if your stack is mainstream SaaS and you want a prebuilt block for everything, Make's breadth wins. If you hit custom APIs, internal services, or want to drop into JavaScript mid-flow, n8n's code-first posture is the better fit.

Who fits which

Choose Make if:

• You run low-to-moderate volume with many tiny workflows — per-credit stays cheap and predictable.
• You never want to manage a server. Zero ops is the whole point.
• You need the widest catalog of prebuilt app connectors out of the box.

Check current Make pricing tiers (affiliate link — see disclosure above).

Choose n8n self-hosted if:

• You're at high volume (tens of thousands of runs/mo and up) and can keep a VPS alive. Unmetered executions make the cost curve nearly flat.
• You want open-source control, data on your own infra, and code-level flexibility.
• You (or someone on the team) are comfortable with Postgres + Redis once queue mode arrives.

Choose n8n Cloud if:

• You want n8n's flexibility but not the ops — and your volume is low enough that per-execution pricing doesn't bite. Avoid it for high-frequency small workflows; that's its worst-cost zone.

Start an n8n Cloud trial (affiliate link — see disclosure above). Note: only n8n Cloud is monetized here; self-hosting n8n is free and we earn nothing on it — we recommend it anyway where the math favors it.

What real users say

ALL workflows fail to execute when n8n is running in queue mode (N8N_EXECUTIONS_MODE=queue). The error occurs immediately when any workflow is triggered (webhooks, manual execution, scheduled triggers, etc.), preventing the execution from being queued and processed by worker containers.

Add Redis Keep-Alive Configuration for Queue Mode

FAQ

Can you self-host Make like n8n?

No. Make is cloud-only and cannot be self-hosted; it is cloud-first at all subscription levels (source: make.com/en/compare/make-vs-n8n, 2026-05-30). Only n8n offers a self-hostable open-source edition with unlimited executions on your own server. If self-hosting is a hard requirement, n8n is the only option of the two.

Why is n8n Cloud more expensive than Make at the same volume?

Because they bill on different units. Make charges per credit (a module/step usually consumes at least one credit), and many tiny workflows keep that count low. n8n Cloud charges per execution (one whole workflow run, regardless of node count), and its tiers are priced higher: Pro is EUR 50/mo for 10k executions versus Make Core at $9 for 10k credits (sources: n8n.io/pricing, make.com/en/pricing, 2026-05-30). n8n's cost advantage only appears when you self-host, where executions are unmetered.

At what volume does n8n self-host beat Make?

Roughly when your monthly run count climbs into the tens of thousands and stays there. At 50k runs, Make rounds up to its 80k slider tier (~$29-43/mo) while a self-hosted n8n CX22 still costs ~$5-9/mo because executions aren't metered (sources: make.com/en/pricing; github.com/n8n-io/n8n, 2026-05-30). At 500k the gap widens further — but only if you can run the server.

How much maintenance does self-hosting n8n actually take?

At low volume, a single-container CX22 is light (~1-2 planning hrs/mo). At high volume you move to queue mode, which means running Postgres + Redis + main + worker processes. Documented failure modes include stuck executions and Redis idle-timeout drops (GitHub issues 17395 and 21319). Budget ~2-3 hrs/mo and real ops familiarity. Maintenance-hour figures are planning estimates, not a single cited source.

Which has more integrations, n8n or Make?

Make states 3,000+ app integrations; n8n states 1,200+ pre-built apps (source: make.com/en/compare/make-vs-n8n, 2026-05-30). Make is broader on prebuilt connectors; n8n leans on code and raw HTTP nodes for flexibility, so technical teams can often reach anything with an API even without a prebuilt app.