n8n vs Make: the short answer
n8n bills you for one run of an entire workflow. Make bills you for every step inside a scenario. That single difference in the pricing model decides most n8n vs Make arguments before any feature comparison starts, because a twenty-step process costs twenty times more to run on one platform than the other.
Beyond billing, the split is clean. Make is the better visual automation tool: a friendlier canvas, more connectors, and a data mapper a non-developer can actually operate. n8n is the better engineering tool: you can self-host it, drop into JavaScript or Python mid-workflow, and build complex workflows that would need awkward workarounds elsewhere.
Neither is a Zapier replacement in the marketing sense — all three automate the same category of work — but Make and n8n both give you far more control over branching, data transformation and error handling than a linear trigger-action tool does. Pick Make if the people building the automations are marketers and ops staff. Pick n8n if the people building them write code, or if the data cannot leave your infrastructure.
Every price and limit below was checked against the vendors' own pages on September 3, 2026. Both companies change plans often, so confirm the current pricing page before you commit budget.
n8n vs Make at a glance
Make at a glance: a hosted, visual automation platform built for operations and marketing teams. Make offers a drag-and-drop scenario canvas, more than 3,000 app integrations, per-step credit billing and a free tier you can build on. There is no self-hosted edition and no way to run it inside your own network.
n8n at a glance: a source-available workflow automation platform built for technical teams. n8n offers a node-graph editor, roughly 2,100 integrations, per-execution billing, JavaScript and Python nodes with no plan gate, and the option to deploy n8n on your own server under a free community licence.
Unlike Zapier, which is designed around one trigger and a linear chain of actions, tools like Make and n8n treat branching, looping, sub-workflows and error paths as first-class parts of the builder. Both will happily run an automation with forty steps, several conditional routes and a retry policy.
The pricing structure is the fault line. Make charges per operation, so cost tracks the number of modules that fire. n8n charges per execution, so cost tracks how often a workflow runs and ignores how big it is. Everything else in this comparison is a preference; that one is arithmetic.
How each platform actually works
Make calls a workflow a scenario, and builds it on a circular canvas of modules connected by lines. Data moves between modules as bundles, and you map fields by dragging them from an upstream module's output into a downstream module's input. Routers split the path, iterators break arrays apart, aggregators put them back together. It is the most legible visual builder in this category once a workflow gets past ten steps.
n8n calls a workflow a workflow, and lays it out as a left-to-right node graph. Every node receives an array of items and emits an array of items, which sounds academic until you hit your first loop: n8n runs most nodes once per item automatically, so what would be an explicit iterator in Make is often just the default behaviour. When the built-in nodes run out, the Code node takes JavaScript or Python and the HTTP Request node takes anything with a REST endpoint.
The practical consequence is who can maintain what. A Make scenario is readable by whoever inherits it. An n8n workflow with three Code nodes in it is readable by a developer. That is not a criticism of either — it is the main thing to weigh when you decide which workflow automation tool your team will still be running in two years.
Pricing model: credits per step vs executions per workflow
This is the section that changes budgets. Make bills in credits, and most standard modules consume one credit each time they run — read a record, create a record, transform a value, each is a credit. n8n charges per execution, and its pricing page is explicit that an execution is a single run of an entire workflow, no matter how many steps are in it or how much data it processes.
Run the same job through both and the arithmetic diverges fast. A scenario that fetches a lead, enriches it, checks a duplicate, writes to a CRM, sends a Slack message and logs a row burns roughly six credits per lead on Make. The equivalent n8n workflow burns one execution. At a thousand leads a month that is 6,000 credits against 1,000 executions.
Make's published plans, checked on September 3, 2026: a Free tier at $0 with 1,000 credits a month, two active scenarios and a 15-minute minimum scheduling interval; Core at $12 a month with 10,000 credits, unlimited active scenarios and one-minute scheduling; Pro at $21 with the same 10,000 credits plus custom variables, full-text search of the execution log and priority execution; Teams at $38 with roles and shared templates; Enterprise on quote. Those figures are the annual-billing rate — monthly billing costs more.
n8n's cloud plans on the same date: Starter at €20 a month for 2,500 executions with unlimited steps, Pro at €50 for 10,000 executions, Business at €667 for 40,000 executions plus SSO and a self-hosted option, and Enterprise on quote, also annual rates. Note the currency difference, and note that n8n includes unlimited users on Starter and Pro while Make gates team roles behind its Teams plan.
Read those two paragraphs together and the pattern is clear. Make is cheaper to start and gets expensive as workflows get longer. n8n costs more per plan at the bottom and stays flat as workflows get longer. Where your business lands depends on step count per run far more than on run count. Our breakdown of what n8n really costs on cloud versus self-hosted works that maths through in detail.
Self-hosting: n8n's structural advantage
Make is a hosted product. There is no version you run on your own server; the closest thing is an on-premise agent on the Enterprise plan that reaches into your network on the platform's behalf. If your data cannot leave a vendor's cloud, Make is out of the conversation.
n8n can be self-hosted, which is the single biggest reason technical teams choose it. The Community edition runs on your own hardware at no licence cost under n8n's Sustainable Use License — a fair-code, source-available licence rather than an OSI-approved open source one, which matters mainly because it forbids reselling n8n as a hosted service. Internal business use is explicitly fine.
Self-hosting n8n is not free, it is just differently priced: you are trading a subscription for a server, a database, backups, upgrades and someone to own them. And several features stay behind the commercial licence — SSO, environments, external secrets, log streaming, projects and Git-based version control are all on the paid tiers, while registering a free Community instance unlocks folders and in-editor debugging. If you want the full picture on running AI workloads on your own infrastructure, our self-hosted AI guide covers the hardware and operational side.
Integrations: app coverage vs building your own
Make wins on raw count. Make provides more than 3,000 app integrations according to its own directory when we checked on September 3, 2026, and the long tail is genuinely long — regional payment processors, niche e-commerce platforms, the CRM your industry uses and nobody else does.
n8n listed 2,131 integrations on the same day. The gap is real but less decisive than it looks, because the two platforms treat missing connectors differently. Both ship a generic HTTP request node, so anything with a documented REST API is reachable from either. n8n goes further: community nodes install from npm, and writing your own node is a normal thing for a developer to do in an afternoon.
So the honest test is not the total. It is whether the specific five apps your business runs on are supported natively, and how ugly the fallback looks. Check both directories against your actual stack before you decide — a missing connector for your accounting system costs more than a thousand connectors you will never open.
Data transformation, error handling and complex workflows
Make's data transformation is visual and, for the common cases, better. The mapping panel shows what each field will contain, built-in functions cover text, date and array work, and the iterator/aggregator pair handles most array reshaping without code. Custom JavaScript functions, though, are an Enterprise feature — on lower plans you work within the built-in function set.
n8n's model is code-first when you need it. The Code node runs JavaScript or Python over the whole item array or item by item, expressions embed JavaScript inline in any field, and there is no plan gate on either. For genuinely complex workflows — recursive API pagination, conditional retries with backoff, reshaping deeply nested JSON — this is the difference between an elegant workflow and a scenario held together with sub-scenarios.
On error handling both platforms are competent and differently shaped. Make attaches error handler routes to individual modules, with directives that resume, ignore, break or roll back, and error handler modules do not consume credits. n8n uses a dedicated error workflow that fires whenever a workflow fails, plus per-node retry settings and continue-on-fail, which centralises alerting nicely across dozens of workflows. If you run automations that must not silently die, test this area before you buy — it is the part every comparison skips and every operator lives in.
AI features: agents, models and who pays for tokens
Both platforms now ship AI agents rather than just an OpenAI connector, and both put the model behind their own billing by default.
Make's AI agent modules cost one credit per operation plus additional credits for the AI tokens when you use Make's own AI provider. Connect a custom provider — an OpenAI or Anthropic account of your own, which paid plans allow — and Make charges the operation credit while the model provider bills you directly for tokens. Uploading knowledge files carries its own credit and token cost. That is worth modelling before you scale an agent, because AI modules are exactly the ones that consume more than one credit per run.
n8n's AI nodes are built on the LangChain pattern: an agent node wired to a chat model, tools, memory and a vector store, each as its own node on the canvas. Because you supply the model credential, model spend goes straight to the provider, and a self-hosted n8n instance can point at a local model with no third-party call at all. We walk through that build in detail in our guide to building AI agents in n8n.
For most businesses the decision here is not which vendor's AI is smarter — it is who holds the API key. Bundled AI usage is simpler and marked up; bring-your-own-key is cheaper at volume and gives you model choice. We compare the two approaches in our guide to owning your API keys.
Where Make is the right answer
Choose Make when the people building automations are not developers. The canvas is more forgiving, the field mapper explains itself, and an ops manager can debug a broken scenario without opening a terminal. That advantage compounds — automations built by the team that owns the process get maintained; automations built by a contractor in a tool nobody else understands rot.
Choose Make when your workflows are short and numerous. Its credit model punishes long scenarios and rewards small ones, so hundreds of three-step automations are genuinely cheap here. The Free tier's 1,000 credits are enough to prove a use case before spending anything.
Choose Make when you need an obscure connector it already has and n8n does not, when a full-text searchable execution log matters for support work, or when you simply have no appetite for hosting anything. A hosted platform where somebody else owns uptime is a legitimate answer, not a compromise.
Where n8n is the right answer
Choose n8n when workflows are long. Per-execution pricing means a forty-step process costs the same as a four-step one, and that is the fastest way to make an automation platform bill stop growing with your ambition.
Choose n8n when the work is technical: custom API calls, real branching logic, data reshaping that needs a language rather than a mapping UI, or AI agents wired to your own tools. Anything you can write in JavaScript, you can run mid-workflow without leaving the canvas.
Choose self-hosted n8n when data residency, compliance or client confidentiality mean records cannot sit in a third party's cloud. Agencies handling client credentials, clinics handling patient data and firms with contractual data-location requirements all land here, and it is usually a hard requirement rather than a preference.
And choose n8n when you want the option to leave. Workflows are JSON you can export, the source is available, and the same instance runs on your laptop or your own server. n8n could be more polished; what it offers instead is that you own the deployment.
Where Zapier still fits
Most people reading an n8n vs Make comparison arrived from Zapier, so it is worth being straight about it. Zapier is the easiest of the three to start with and the most expensive to scale, and it dominates the shallow end for a reason: the largest app directory, the best templates, and setup measured in minutes. Zapier excels at exactly one thing — getting a working automation live before lunch.
Zapier charges per task, and its own documentation counts a task as a successfully completed unit of work while excluding triggers, filters, paths and formatter steps (checked September 3, 2026). Its cost curve therefore looks more like Make's than n8n's — Zapier and Make both meter steps, n8n meters the run. Compared to Zapier, Make gives you real branching and loops for less money, and n8n gives you code and self-hosting. Zapier allows multi-step paths and ships its own AI features, which has narrowed the functional gap, but the pricing gap at volume has not closed.
Zapier is still the right pick in two situations. First, when a non-technical owner is automating a handful of processes and will never build more — paying a premium to never think about it is rational. Second, when Zapier has the only native connector for a critical app, which happens more often than the other two vendors like to admit.
If you want the three-way rather than this head-to-head, Zapier vs Make vs n8n is a separate comparison with a different conclusion at each business size. The n8n vs Zapier and Make vs Zapier questions both resolve the same way in the end: Zapier for simplicity, Make for visual power at mid volume, n8n for control and cost at high volume.
Migrating between them
There is no reliable one-click converter between these platforms. Switching from Zapier or Make to n8n means rebuilding, and the honest estimate is that effort scales with step count and connector obscurity, not with the number of workflows.
Rebuild in three passes. First, inventory what actually runs — most organisations find a third of their automations are dead, duplicated or triggered by nothing. Second, port the high-volume workflows that automate work every day, since those are where per-step billing was hurting and where the payback sits. Third, leave the odd one-off automations on the old platform until their connectors exist on the new one; running both for a quarter is cheaper than a rushed migration.
Two traps. Credentials must be reissued in the new platform rather than exported, so budget time for OAuth reconnection across every app. And behaviour differs subtly around empty results, pagination and time zones, so run both versions in parallel against real data before you switch the trigger off.
Use cases: which platform is best suited to which job
Marketing and sales glue — a form fill that enriches a lead, writes to the CRM, posts to Slack and starts an email sequence — is where Make is best suited. Make supports the relevant apps natively, the scenario stays readable, and at four or five modules per run the credit cost stays small.
E-commerce order and inventory sync across a storefront, a warehouse system and an accounting package also favours Make, purely on connector coverage. Check both directories, but the long tail of shopping-cart and payment integrations is deeper there.
Document and invoice processing with AI extraction can go either way. Make AI modules make the build faster; n8n allows you to swap models freely and keeps token spend on your own provider account. Step count usually settles it, because document pipelines tend to be long.
Internal API work — recursive pagination, syncing two databases, reshaping nested JSON, glueing a legacy system to a modern one — belongs on n8n. So does anything client-confidential, since n8n workflows can run entirely inside your own network.
Long operational workflows are the clearest case of all. Once a process runs thirty or forty steps per execution, Zapier and Make both bill for every one of them and n8n does not. That is the point at which most teams that automate seriously start looking at a migration.
Which should your business choose?
Choose the right platform by answering three questions, not by reading feature tables. How many steps does a typical workflow contain? Who maintains the automation once it is live? Where must the data live?
Long workflows, a developer on the team, or a data-residency requirement point to n8n. Short workflows, a non-technical builder and no appetite for infrastructure point to Make. With fewer than ten simple automations and no in-house technical owner, choose Zapier and feel no shame about it.
The failure mode we see most often is not picking wrong — it is picking a platform before mapping the process. A badly designed workflow costs more on every tool. If you would rather start from the process than the software, our free automation audit takes about three minutes, needs no call, and returns an automation health score, the annual cost of the manual work you describe and a ranked list of quick wins. If you would rather someone else build it, our AI agent builds start at $7,500 and typically ship in about 30 days.
Frequently asked questions
Which is better, n8n or Make?
Neither is better across the board. n8n is better for long workflows, technical builders and anyone who needs self-hosting, because it charges per workflow run and lets you write code. Make is better for non-technical teams, short scenarios and obscure connectors, because its visual builder is friendlier and its app directory is larger. Teams running more than a handful of automations usually end up sorting workflows by step count and moving the long ones to n8n.
Is Make cheaper than n8n?
At low volume with short scenarios, usually yes — Make's entry plans start lower and its free tier is usable. As soon as workflows get long, Make's per-step credits overtake n8n's per-execution pricing. Count the steps in your most frequent workflow and multiply by monthly runs; that number tells you whether Make or n8n is the cheaper automation platform for your actual usage.
Do you need development skills to use n8n or Make?
Not to start with either. Make is genuinely usable by non-developers for most tasks. n8n is approachable for simple workflows but rewards someone who can read JSON and write a little JavaScript, and self-hosting definitely needs technical ownership.
Can n8n handle enterprise-scale workloads?
Yes, with the right deployment. Its Business and Enterprise tiers add higher concurrency, environments, SSO and Git-based version control, and self-hosted instances scale out across worker processes. The constraint is usually your own operations capacity rather than the software.
Can Make be self-hosted?
No. Make is a cloud platform with no self-hosted edition; its Enterprise on-premise agent connects to systems inside your network but the platform itself still runs in Make's cloud. If you need the automation engine on your own servers, n8n is the choice among these two.
Is there a better alternative to n8n or Make?
It depends on what is missing. Zapier if you want maximum simplicity and connector coverage, Power Automate if you are deep in Microsoft 365, or a custom-built service if the workflow is core to your product and you have outgrown visual tools like n8n and Make. Most businesses do not need a fourth option — they need to finish designing the process.
How hard is it to migrate from Make to n8n?
Plan on a rebuild rather than an import. Workflow logic ports quickly for anyone who knows both tools; credentials must be reconnected app by app, and edge-case behaviour differs, so run the old and new versions side by side against live data before cutting over.