Credits vs executions, cloud vs your own server, and the license question
Make and n8n both draw workflows on a canvas, but Make bills per module run and n8n bills per workflow run or runs free on your own server. Here is how to choose between them.
At a glance Make and n8n look alike. Both show a workflow as boxes joined by lines and both handle branching and loops. What separates them is where the software runs, how you pay for it, and how much of the plumbing you are willing to own.
Make is a hosted platform. You build a scenario on the canvas and Make runs it on its servers. It lists 3,500+ applications, and its execution view shows how many data bundles passed through each module, which makes debugging concrete.
n8n comes two ways. n8n Cloud is hosted by the vendor. The Community edition is a build you run yourself with Docker, and n8n's docs say it includes "almost the complete feature set." n8n lists 2,000+ integrations, and when an app is missing you reach for the HTTP Request node or a small code node. n8n expects you to be comfortable with an API doc now and then.
Make bills in credits. Its pricing page says most module actions cost one credit, and routers and error handlers cost nothing. The expensive part is volume inside a run: a trigger that returns ten records makes every later module run ten times, and Make's own help example shows a single run costing 31 operations. The Free plan includes 1,000 credits a month (checked November 2026).
n8n Cloud bills per execution, which the pricing page defines as "a single run of your entire workflow," and it states that pricing is based on executions "regardless of complexity." A 30-step workflow that runs once is one execution. The Starter cloud plan lists 2,500 executions a month at EUR 20 a month billed annually (checked November 2026). There is a free trial but no permanent free cloud tier.
Self-hosting changes the math again. The Community edition costs you a server and your time, and executions are not metered.
n8n is not open source in the usual sense. Its code is under the Sustainable Use License, which faircode.io lists as a fair-code license. The license permits use "only for your own internal business purposes or for non-commercial or personal use," and lets you provide it to others "only if you do so free of charge for non-commercial purposes." Automating your own company is fine. Reselling n8n as a hosted service to clients is not covered. Make is proprietary software you rent, so the question does not arise.
Make AI Agents are, in Make's words, "available on all plans" and "work alongside deterministic logic, not instead of it." n8n includes its AI Agent node on all cloud plans, with human approval for tool calls, and its docs explain that the AI features implement LangChain concepts, connecting models from OpenAI, Anthropic and Google in one workflow. n8n gives a developer more control over memory and tools. Make gives a non-technical builder a shorter path to a first agent.
Make is the gentler start. n8n's editor is comparable once you are in it, but self-hosting adds Docker Compose (the option n8n's hosting docs list for production deployments), backups, updates and credential storage to your plate. If nobody on your team has run a server, that is a real cost.
A team with no engineer should skip self-hosting. Pick Make if runs are small and branching is complex, n8n Cloud if runs process many records at once.
A developer who wants to self-host, keep customer data inside the company network and avoid a per-run bill should pick n8n Community. Check the license fits how you plan to use it.
A marketer with a CRM and no server access should pick Make and watch the bundle count.
The AI Automation with Zapier, Make and n8n course has you build one Tally-to-HubSpot lead intake on Zapier, Make and n8n. In chapter 2 you compare what Make's visual bubbles tell you against what an execution log tells you. Chapter 3 opens the HubSpot record each run wrote and checks every field. Chapter 5 adds one AI step with a price, a parser and a daily cap, the same discipline whether the model step sits in a Make module or an n8n node.