n8n vs. Make: Operation Billing vs. Flat Hosting

Make (formerly Integromat) and n8n are the two most popular tools for automating workflows that go beyond simple two-app triggers. They look similar on the surface — visual editors, hundreds of connectors, webhook support. But their pricing models are structurally different, and that difference compounds fast once you add AI nodes, multi-step sequences, or decent volume.

This post runs the two models side-by-side on a concrete scenario so you can see where Make's operation counter bites, and what a flat-priced n8n instance changes.

---

How Make's Pricing Actually Works

Make sells operations, not workflows or automation minutes. An operation is counted every time a module executes — every step in a scenario that processes one item.

So if your scenario has 6 modules and it runs on 100 items:

6 modules × 100 items = 600 operations

Make's Core plan gives you 10,000 operations/month. Sounds generous until you add two things:

1. Multi-step scenarios — a realistic flow (trigger → filter → transform → HTTP call → AI node → send notification) is already 6+ modules.
2. AI / LLM calls — Make's OpenAI and Claude modules each count as a module execution. The AI tokens themselves are billed separately via your own API key. Make does not include AI credits. You bring your own key and pay OpenAI or Anthropic on top of your Make subscription.

The Operation Counter in Practice

Take a lead enrichment scenario that runs every hour:

Step	Module
1	Schedule Trigger
2	HTTP Request (fetch new leads)
3	Iterator (split array into items)
4	HTTP Request (enrich each lead via an external API)
5	OpenAI — generate a personalized outreach draft
6	IF (filter low-quality leads)
7	Gmail — send the draft

That's 7 modules per lead. Process 200 leads/day → 1,400 ops/day → 42,000 ops/month. You blow through the Core plan (10,000 ops) in under 8 days and land on a higher paid tier before the month ends.

Add a second scenario — say a Slack digest that summarizes customer tickets with AI — and you're stacking another 15,000–20,000 ops on top.

The structural problem: Make's meter ticks for every item × every module. More complex flows and higher volume push you into higher tiers quickly, and you're still paying OpenAI separately for every AI call.

---

How n8n Pricing Works

n8n's cloud pricing is also per-execution (they call them "workflow executions"), but the unit is the full workflow run, not each node inside it. The more important option for power users is self-hosting n8n — running your own instance on a server you control, where there's no per-execution charge at all. You pay for the server; the workflow runs are unlimited.

This is the model that changes the math.

On self-hosted n8n:

• No operation counter.
• No execution limit.
• You configure your own credentials (OpenAI, Anthropic, etc.).
• You're responsible for the server, SSL, updates, backups, and uptime.

That last point is where the actual work lives, and it's not trivial — which leads to the third option.

---

The Honest Comparison Table

	Make Core	Make Pro	n8n Cloud Starter	Your Own n8n (self-hosted)
Monthly price	~$10.59	~$16	~$20	$5–20 server + your time
Execution model	Per operation	Per operation	Per execution	Unlimited
AI credits included	No (BYOK)	No (BYOK)	No (BYOK)	No (BYOK)
DevOps work	None	None	None	High — you manage everything
Data ownership	Make's cloud	Make's cloud	n8n's cloud	Your server
Scales linearly with volume?	Costs scale up	Costs scale up	Costs scale up	Flat

Notice what every column in that table has in common: no AI credits included. Whether you're on Make, n8n Cloud, or a $10 VPS running n8n yourself, AI nodes require your own API key and bill separately.

---

Who Should Pick What

Make is the right call if:

• You're running simple, low-step scenarios at low volume.
• You want the fastest no-code setup with zero server management.
• You don't use AI nodes much (or at all).
• Your workflows run maybe a few hundred operations a day.

n8n self-hosted is the right call if:

• You need unlimited runs and don't want an operation counter.
• You have the engineering time to manage Docker, nginx, SSL certs, and updates.
• You want full data sovereignty.

n8n Cloud is the right call if:

• You want n8n's visual editor without the server work.
• Per-execution pricing works for your volume.
• You're OK managing API keys for AI nodes.

---

Where the AI Cost Reality Bites Everyone

Here is the hidden cost that both Make and n8n Cloud users discover too late: AI nodes are always bring-your-own-key.

That's not a criticism of either product — it's just how they're built. But it means:

• Your OpenAI API bill is separate and unbounded.
• You need to monitor token usage yourself.
• AI-heavy scenarios (summarization, classification, outreach drafts, data extraction) generate API bills that can easily exceed your Make or n8n Cloud subscription cost within weeks.

For any workflow that calls an LLM on each processed item — think lead scoring, document extraction, or email drafting — that external API bill scales directly with your volume. The more useful your automation becomes, the more it costs in AI tokens, billed outside the platform entirely.

The standard advice is to switch to cheaper models when costs climb. That's valid, but it means you're constantly managing which model to use for cost reasons rather than capability reasons.

---

How to Do This on AgentRoost

AgentRoost takes the self-hosted n8n path and removes the DevOps part. When you sign up and pick the n8n framework, you get your own n8n instance — your login, your data, your workflows — running on a public HTTPS subdomain (https://<your-id>.agentroost.app). It is not a shared multi-tenant service. You own the instance.

The key difference from vanilla self-hosting: AI credits are included in the subscription price. The AI/LLM nodes in your n8n instance are already wired to a credit pool. You don't enter an OpenAI API key. You don't get a separate AI bill. You pick from 350+ models — GPT-4o, Claude Sonnet, Llama, Mistral, and more — and switch anytime. The credits are already paid for.

The actual journey:

1. Sign up at AgentRoost with email/password or Google/Microsoft/Discord.
2. Pick the n8n framework and name your instance.
3. Your private n8n editor opens at https://<your-id>.agentroost.app.
4. Build the lead enrichment scenario above. Drop in the AI/LLM node. It works immediately — no API key to enter.
5. Add a Schedule Trigger, point the HTTP Request nodes at your sources, wire the IF node for filtering. Webhooks automatically get a public HTTPS URL. Deploy and walk away.

Pricing starts at $19.99/month all-in — that covers the compute, the SSL, the updates, and the AI credits. There's no operation counter. A scenario that runs 42,000 operations a month costs the same as one that runs 1,000. The 14-day money-back guarantee means you can validate your actual workflow before committing.

Compare plans

---

Practical Node-Level Tips for the Scenario Above

If you're building that lead enrichment flow (in n8n, on any hosting):

• Schedule Trigger — set to 0 * * * * (top of every hour) in cron expression mode.
• HTTP Request node — use "Authentication: Header Auth" for APIs that need a bearer token; store it as a credential, not inline.
• Split In Batches — if your lead list is large (>100 items), batch before the AI node to avoid timeouts.
• AI/LLM node — for outreach drafts, a system prompt like "You are a B2B sales assistant. Write a 3-sentence personalized email opening based on the lead data below. Be specific, not generic." produces better output than a vague instruction. Feed {{ $json.company }}, {{ $json.title }}, {{ $json.recent_activity }} as variables.
• IF node — filter leads where {{ $json.score }} is below a threshold before sending to Gmail. This cuts unnecessary AI calls and email volume.
• Error Workflow — set a dedicated error workflow under Settings → Error Workflow. Route failures to a Slack message node so silent failures don't go unnoticed.

---

The Bottom Line

Make's operation-based pricing is predictable for simple, low-volume scenarios and expensive for multi-step, AI-heavy flows at any real volume. n8n self-hosted removes the counter but adds DevOps burden — and still leaves the AI bill unbounded.

The case for a flat-priced, self-owned n8n instance with AI credits already included is straightforward: one line item instead of three (subscription + server + AI API), no surprises at the end of the month, and the full power of n8n's node library with no per-run meter ticking in the background.