n8n vs Make (2026): self-hosting and credit costs

n8n is an open-source workflow automation tool you can self-host on your own infrastructure or run on n8n Cloud. It is built for technical teams who need complex logic, AI agents, and control over where their data runs.

Make, formerly Integromat, is a cloud-only visual platform with a drag-and-drop scenario builder and a catalog of more than 3,000 apps. It is built for people who want a polished automation working today without touching a server.

They differ on more than the interface. Only one of them can run inside your own network, and the two do not even bill in the same currency.

At a glance: n8n vs Make

Sources: n8n, Make, both checked 07 Aug 2026.

What is n8n? What is Make (formerly Integromat)?

n8n is a workflow automation tool whose source code is published on GitHub under a fair-code license. You can install it on your own server with Docker, or let n8n run it for you on n8n Cloud. The official catalog lists 2,003 integrations, and anything missing from it can still be reached through the HTTP Request node, which calls the API directly.

The editor is a canvas you drop nodes onto and wire together. It shows the data passing through each node as you build, so you read the JSON at every step instead of guessing at it.

Make is a visual automation platform. It launched as Integromat and carried that name for years before the rebrand, and integromat.com still redirects to Make today. So when an older tutorial or an internal runbook says Integromat, it means this product.

There is no self-hosted edition and there never has been. The catalog covers more than 3,000 apps, and you build workflows, which Make calls scenarios, by dragging modules onto a canvas and connecting them in sequence.

When to choose n8n

n8n earns its complexity in four situations.

• Self-hosting and data privacy. You operate in a regulated industry, your data cannot sit on someone else's cloud, and self-hosting (Docker, reverse proxy, backups) is on the table. Make is cloud-only with no self-host option, so n8n is the only credible choice when on-premise execution is a hard requirement.
• Complex logic and transparency. You need deep branching, loops, error handling, custom JavaScript functions, and JSON payloads you can inspect rather than a "black box" execution engine. n8n exposes the data structure at every node; Make abstracts it behind the visual canvas.
• AI agents with model choice. You want to wire workflows to your own LLMs (Claude, OpenAI, Mistral, self-hosted Llama), control prompts at the node level, and inspect token usage. n8n's LangChain-native nodes make this first-class; Make's AI offering is more constrained.
• Cost efficiency at high volume. You run tens of thousands or millions of actions a month and you have the technical resources to keep a server alive. Past a certain volume, the flat cost of a self-hosted n8n box comes in under the cost of buying more Make credits.

Where n8n hits its limit

Self-hosting is real work. "Free" n8n is free software, not free infrastructure. You pay for the server (typically $5 to $15 per month on a small VPS), and you pay in your own time for setup, Docker, reverse-proxy config, backups, version upgrades, and the occasional 11pm "why did this workflow stop running" debugging session.

The learning curve is real too. n8n exposes expressions in a JavaScript-ish syntax, raw API payloads, and node wiring patterns that take days to internalize. A non-technical user without an automation background will ship their first useful scenario on Make weeks earlier than their first workflow on n8n. The payoff is unmatched control. The cost is the runway.

When to choose Make

Make wins in four situations.

• Speed of build. Make's scenario builder is the most polished visual canvas in the category. You can wire eight modules together, watch data flow through each one, and ship a working scenario faster than on n8n's node editor.
• Zero infrastructure. You want fully-managed cloud automation with SOC 2 controls, zero server upkeep, and a platform that handles uptime, retries, and observability for you. n8n Cloud also does this; Make does it with more polish for non-technical operators.
• Pre-built modules for standard SaaS. Make's catalog covers most common SaaS tools out of the box (Notion, Google Workspace, Slack, Airtable, HubSpot) with rich modules that expose the specific actions you actually use, not just generic webhooks.
• Low-to-mid volume affordability. Make's credit pricing starts at $12 per month billed monthly for 10,000 credits on the Core plan, or $9 per month if you pay for the year, and it scales gently for short sequential workflows. At small scale you are often cheaper on Make than on n8n Cloud.

Where Make hits its limit

No self-hosting, ever. Make is cloud-only. If your industry, security policy, or compliance posture requires on-premise or private-cloud execution, Make is off the table. n8n is the only credible alternative for that constraint.

Credit pricing compounds at scale. Make bills in credits, and most module actions cost one credit. A scenario with 8 modules running 1,000 times consumes 8,000 credits. At low volume that is cheap. At high volume the bill climbs fast, and once you are buying credits by the million the flat infrastructure cost of self-hosted n8n comes in an order of magnitude lower.

Limited AI control. Make's AI modules and AI Agent builder work for straightforward use cases (summarize this, classify that, route based on intent) but constrain prompt and model choice. If you want to swap LLMs, run agents on self-hosted Llama, or inspect token usage at the node level, n8n wins on depth.

Pricing and total cost of ownership

To make the comparison concrete, here is what each option costs for a typical mid-volume workload: a calendar-to-database sync running roughly 200 events per month in two directions with 8 mapped fields, which works out to about 5,000 credits on Make and about 400 workflow executions on n8n.

The two vendors do not bill the same way. Make charges in US dollars, and the $12 above is the monthly-billing price ($9 per month if you pay annually). n8n charges in euros, and the €20 above is the annual-billing price, so paying monthly costs more (annual billing carries a 17% discount). Different currency, different cycle: the two figures cannot be read as a straight cheaper-or-dearer comparison. Both come from the vendors' own pricing pages, checked 07 Aug 2026.

At this volume, the three options break down like this.

• Make Core is the cheapest subscription at $12 per month billed monthly for 10,000 credits, or $9 per month on the annual plan. The visual builder is also the fastest to ship into, so for low-to-mid volume with apps already in Make's catalog it wins on time to value.
• n8n Cloud Starter is €20 per month billed annually for 2,500 executions on a more technical interface, and Pro is €50 per month billed annually for 10,000 executions. Because n8n bills in euros, what lands on a dollar card depends on the rate your issuer applies that day. The depth of control earns the price if you need branching, JavaScript injection, or LangChain workflows.
• Self-hosted n8n is cheapest on subscription and expensive in your time. A Notion user without DevOps capacity will spend more on setup, upgrades, and 11pm "why did this workflow stop running" debugging than a managed subscription costs.
• Moving up a Make tier does not buy more credits. Pro ($21 per month billed monthly, $16 annual) and Teams ($38 per month billed monthly, $29 annual) both list the same 10,000 credits as Core. What the higher tiers add is priority execution and team features; more throughput means buying a larger credit allowance inside whichever plan you are on. So the credit bill keeps climbing with volume, and past a certain point the flat server fee of a self-hosted box comes in cheaper.

AI agents in 2026

Both n8n and Make have invested heavily in AI workflows over the last 18 months. The gap between their approaches is widening rather than closing.

• n8n is the builder's choice. LangChain-native nodes give you the same primitives developers use to ship production AI agents elsewhere. You pick your model (OpenAI, Anthropic, Mistral, self-hosted Llama), you control the prompt at the node level, you inspect token usage, and you can self-host the whole stack including the LLM. For agents that operate across many tools with custom logic, n8n is the stronger fit in 2026.
• Make is the operator's choice. The AI Agent builder sits inside the same drag-and-drop canvas non-technical users already know. On the ecosystem side, Make ships connectors for the major AI services (OpenAI, Anthropic Claude, Google Gemini, Perplexity), and the Make MCP server lets an outside assistant such as ChatGPT or Claude call your scenarios directly. You never think about LangChain or model parameters; you describe what the agent should do and Make handles the orchestration. For agents a non-technical operator has to maintain, Make wins on UX.

Neither tool, however, is built to keep your underlying data consistent between systems on its own. That is a separate problem, and the next section covers where it shows up.

Where neither tool fits: continuous two-way sync

Both n8n and Make share a shape: event-triggered automation. Something happens, the tool reacts, the workflow ends. That model fits most jobs cleanly, but it leaks on one class of problem: keeping two systems continuously consistent in both directions, with conflict resolution. The version we run into most often is a Notion workspace paired with a calendar, task list, or inbox, though the limitation is generic. Where a dedicated sync layer typically wins:

• True two-way. Faking a two-way sync between Notion and Google Calendar in Make takes two scenarios, one per direction, with no shared state. An n8n version takes two workflows with state managed manually. Both can race on updates, and both leave duplicates or stale rows when the same item passes through both directions. A sync layer like 2sync handles two-way as a single configuration with conflict resolution built in and a sync interval of 2 to 5 minutes depending on plan. Our step-by-step guide to connecting Notion and Google Calendar walks the setup end to end.
• Field-level control. n8n and Make make you write the mapping logic for each field on each scenario. 2sync maps 16+ fields between Notion and Google Calendar (title, description, dates, attendees, location, recurrence, organizer, color, free/busy, conference link, and more) with a single visual mapping screen, and each field can be set to two-way, one-way to Notion, or one-way to provider.
• Recurring events. n8n and Make both see a recurring Google Calendar event as a single object. They cannot create one Notion page per occurrence without you writing custom logic. 2sync creates one Notion page per occurrence by default, with an Is Recurring property you can filter on. If you want the alternative, native templates that generate future instances inside Notion, see our Notion recurring tasks guide.
• No infrastructure. n8n self-hosted is cheap on paper. In time, it is expensive. A Notion user without DevOps capacity will spend more on n8n maintenance than on a 2sync subscription. 2sync runs at $9/month on the Solo plan, or $7/month billed annually, with zero servers and zero monitoring on your end. The same model covers Todoist, Outlook Calendar, and Gmail without changing your bill.
• No learning curve. No scenarios to build, no nodes to wire, no expressions to write. Pick a Notion database, map fields, go live. The setup wizard runs six steps and takes 5 minutes.

Decision matrix: pick by need, not by brand

The three tools optimize for different problems. The cleanest decision rule is to start from your actual workflow and pick the tool that solves it.

Stacking is fine and common. A typical pattern: 2sync for the calendar, task, and contacts sync, where its two-way handling is strongest, and n8n or Make for the long tail of one-off triggers across other apps. Each tool stays in its lane, and the bill stays lower than putting everything on one platform.

If other tools are on your shortlist, these comparisons cover them:

• Zapier vs Make applies the same cost method with Zapier in the mix.
• Zapier vs n8n puts n8n against the wider Zapier ecosystem.
• Activepieces vs n8n covers the MIT-licensed open-source alternative on license terms and AI agents.
• Zapier alternatives for Notion ranks the rest of the field.
• Best Notion integrations goes past automation to everything that plugs into Notion.

Conclusion

n8n and Make are both real tools that solve real problems, and the choice between them is material: depth and control on one side, speed and polish on the other. Most readers searching this comparison are conflating two questions, though, and the right answer depends on which one they are actually asking. Do you need automation, or do you need sync?

If your problem is wiring many apps together with custom logic, n8n or Make is your tool, and the rest of this article tells you which fits your stack. If your problem is keeping Notion aligned with your calendar, tasks, or inbox, that is a sync-layer problem rather than an automation one, and 2sync was built for it. The 2sync vs Make comparison and 2sync vs n8n comparison lay that out side by side against each tool.

FAQ