n8n vs Make 2026
Short answer: pick n8n if your workflows are long or you need to self-host, pick Make if you want the gentlest no-code start and the biggest app catalog. Same overall score (4.2/5) in our hands-on tests, very different bills.
The catch nobody updated: Make renamed "operations" to "credits" in August 2025, and AI modules now burn a variable, not fixed, amount. n8n still bills one execution per full run, so a 200-step workflow costs the same as a 2-step one. That single difference decides most of this match.
Bills per full run, self-hosts free, deepest AI stack. Technical pick.
Try n8n for free →Read the full n8n review →Gentlest start, 3,000+ apps, but credits burn per step and per poll.
Try Make for free →Read the full Make review →Who wins for you
Make's drag-and-drop canvas and 1,000-credit free plan get you live without touching JSON or Docker.
Try Make for free →Free Community Edition, unlimited executions, JavaScript and Python anywhere. Just patch it: a CVSS 9.9 RCE was exploited in 2026.
Try n8n for free →Per-execution billing means a 20-step scenario costs the same as a 2-step one. Make would charge 20 credits per run.
Try n8n for free →At 1 to 3 steps and high frequency, Make Core near $9 to $12 a month often beats n8n Cloud, unless you self-host.
Try Make for free →n8n vs Make at a glance
Every cell below is grounded in each tool's official pricing and docs as of June 2026. Units differ, so read the billing row first.
| n8n | Make | Edge | |
|---|---|---|---|
| Billing unitThe single biggest difference | 1 execution = 1 full workflow run, any number of steps | 1 credit per module step (renamed from "operations" in Aug 2025) | n8n |
| Entry cloud price | Starter ~€24/mo (€20 annual), 2,500 executions | Core $12/mo (~$9 to $10.59 annual), 10,000 credits | — |
| Free permanent tier | No, 14-day cloud trial only (self-host is free) | Yes, $0 with 1,000 credits/mo and 2 scenarios | Make |
| Self-hosting | Yes, free Community Edition, unlimited runs | No, cloud-only on standard plans | n8n |
| Integrations | ~1,000 to 1,200 (400+ core, 600+ community) plus HTTP node | 3,000+ pre-built apps plus HTTP module | Make |
| AI capabilities | 70+ AI nodes, 12 vector stores, 6 agent types, MCP | Visual AI Agents with Reasoning panel, no native vector stores | n8n |
| Custom code | JavaScript and Python on every plan | JavaScript on Enterprise only | n8n |
| Multiple triggers/workflow | Yes | No, one trigger per scenario | n8n |
| Data residency | Full control self-hosted, or EU Frankfurt on Cloud | EU data centre available, but SCCs may move data outside EEA | n8n |
| Security posture 2026 | Cloud is managed, but self-host hit a CISA-confirmed RCE (CVE-2025-68613) | Cloud-managed, SOC 2 II + ISO 27001, no self-host CVE exposure | Make |
| Community | 30,000+ forum, 80,000+ Discord | 45,000+ community forum | — |
| Ideal user | Developers, agencies, data-sovereign teams | No-code builders, marketers, fast simple automations | — |
Prices checked June 2026 on n8n.io/pricing and make.com/en/pricing. n8n Cloud is priced in EUR, Make in USD, so totals are not directly comparable.
Criterion by criterion, head to head
The same five criteria we scored on each tool's review page. Equal scores still get a clear pick.
01 Round 1: getting the first workflow live.
This one is a genuine draw at 3.8 each, and it surprised us until we watched two juniors learn both side by side. Make wins the first hour: the left-to-right canvas, guided field mapping and inline help got a Slack-to-Airtable scenario running in about 10 minutes, no raw data concepts in sight. n8n's free-form node graph took our junior closer to 3 hours to feel comfortable, because configuring an HTTP Request node means understanding headers, JSON and expressions.
But the curve flips once things get real. n8n shows sample output next to every node as you build, so debugging a 10-node flow is visual and fast. Make's deeper tools (routers, iterators, aggregators) cost a second learning wall, and Capterra reviewers repeatedly flag that error causes are hard to pinpoint. So: easiest day one goes to Make, easiest week two goes to n8n. Net, they tie.
Choose n8n if you will hit routers, custom code and silent-failure debugging fast.
Choose Make if you want a working automation this afternoon with zero code.
02 Round 2: where the bill actually lands.
n8n takes this 4.5 to 4.0, and the reason is structural. It bills per execution: one full run is one execution whether it crosses 2 nodes or 200. Make bills per credit, one per module step, so a 10-step scenario run 1,000 times burns at least 10,000 credits. Worse, polling triggers count credits even when they find nothing. A trigger checking every minute eats roughly 43,200 credits a month before a single useful action fires, and audited accounts have wasted real money this way.
Do the math the dossier supports. A 5-step workflow at 10,000 runs needs 50,000 credits on Make, five times the Core base, while n8n Cloud Pro covers all 10,000 runs at €50 a month flat. Make claws it back at low volume: 5 steps at 2,000 runs is exactly the 10,000-credit Core base near $9 to $12. And self-hosted n8n is simply free software on a €5 to €15 VPS.
Choose n8n if workflows are 5+ steps, high volume, or you can self-host.
Choose Make if you run short scenarios at modest frequency and want a $0 entry.
03 Round 3: raw power and AI depth.
Both land at 4.5, because they are deep in different directions. Make's visual orchestration is genuinely strong: routers, iterators, aggregators, zero-credit error handlers and data stores let you build branching logic without code, and the scenario builder shows data moving between modules. The wall is custom code, JavaScript only lands on Enterprise, and each scenario allows a single trigger.
n8n answers with code anywhere (JavaScript and Python on every plan), multiple triggers per workflow, and an AI stack that is hard to match: 70+ AI nodes on LangChain primitives, 6 agent types, 12 vector stores, 8 memory nodes and native MCP so a workflow can be called by Claude or GPT. Make ships its own AI Agents with a clear Reasoning panel and Module Tools, more approachable, but no native vector stores or RAG layer. For no-code branching, Make. For agentic and RAG systems, n8n. Hence the tie.
Choose n8n if you are building RAG pipelines, AI agents or anything code-heavy.
Choose Make if you want visual branching and reasoning you can read without code.
04 Round 4: who answers when it breaks.
Make edges this 4.0 to 3.5, and it is closer than the numbers suggest. n8n's documentation is thorough and its community is alive (forum answers in 2 to 3 hours, an 80,000-strong Discord), but paid email support tops out around 48 hours with no live chat, and self-hosters are largely on their own through GitHub. When your own server is the product, that gap stings.
Make offers email from paid tiers, a 500+ article knowledge base, decent video tutorials and a 45,000-member forum, and our own three tickets were answered within 6 to 24 hours. The honest caveat: live chat is Teams and Enterprise only, and the harshest community reviews are about support, one user locked out after an account compromise, another whose startup quota burned in two days with no real help. So Make wins on breadth of channels, but neither is flawless when things go truly wrong.
Choose n8n if you are comfortable self-serving via docs, forum and GitHub.
Choose Make if you want a knowledge base, tickets and chat on higher tiers.
05 Round 5: catalog size vs the universal node.
Tied at 4.5, and this is the cleanest tie of the five. On raw count Make wins outright: 3,000+ pre-built app connectors against n8n's roughly 1,000 to 1,200 (400+ core nodes plus 600+ community ones). If your stack is mainstream SaaS, Make almost certainly has a polished native connector waiting.
n8n closes the gap two ways. Its HTTP Request node turns any REST or GraphQL API into a working node (paste a curl command and go), so the practical ceiling is effectively unlimited, and we have wired obscure European tools in under 30 minutes. The bémol cuts both ways: n8n's 600+ community nodes are third-party maintained and quality varies, while some of Make's niche connectors expose only part of the underlying API. Breadth out of the box goes to Make, build-anything flexibility goes to n8n. Even round.
Choose n8n if you hit niche or internal APIs and live in the HTTP node.
Choose Make if you want the widest set of polished native connectors.
The real cost, plan by plan
Two billing models that do not map onto each other. We list the plans, then run the exact cost examples the data supports, assumptions stated.
| n8n | Make | Edge | |
|---|---|---|---|
| FreeMake has a permanent free cloud tier, n8n does not | Self-hosted Community: unlimited runs, you pay €5 to €15/mo for a VPS | $0: 1,000 credits/mo, 2 scenarios, 15-min interval | Make |
| Entry plan | Starter €20/mo annual: 2,500 executions, 5 concurrent | Core $12/mo (~$9 to $10.59 annual): 10,000 credits, unlimited scenarios | — |
| Mid plan | Pro €50/mo annual: 10,000 executions, 20 concurrent, 150 AI credits | Pro $21/mo: 10,000 credits, priority execution, full-text log search | — |
| Team planDifferent scope: n8n Business adds SSO and governance | Business €667/mo annual: 40,000 executions, SSO, Git, 30-day insights | Teams $38/mo: 10,000 credits, team roles, shared templates | Make |
| Enterprise | Custom: unlimited projects, 200+ concurrent, 1,000 AI credits, SLA | Custom: SSO/SCIM, credit overage protection, SOC 2 II + ISO 27001, SLA | — |
| 5 steps × 10,000 runs/moAssumes no polling, no AI modules, standard 1-credit steps | €50/mo on Pro, all 10,000 runs included, steps do not matter | 50,000 credits needed vs 10,000 in Core, ~40,000 in add-ons (price opaque) | n8n |
| 5 steps × 2,000 runs/moMake is cost-competitive at low volume and low step count | €20/mo on Starter, 2,000 of 2,500 executions used | 10,000 credits, exactly the Core base, ~$9 to $12/mo | — |
| 20 steps × 2,000 runs/moStep count is where per-credit billing hurts | €20/mo on Starter, steps still irrelevant to the bill | 40,000 credits vs 10,000 included, overage required, price opaque | n8n |
Prices checked June 2026. Make add-on credit bundles (1,000 or 10,000) are not itemised publicly, you buy them in-dashboard, which is a known UX complaint. n8n Business overage is €4,000 per extra 300,000 executions.
Pick by scenario
Choose n8n if…
- You self-host for data sovereignty (healthcare, finance, public sector) and can own patching
- Your workflows run many steps per execution, so per-run billing crushes the per-credit math
- You are building AI agents or RAG with vector stores, memory and MCP
- You want JavaScript or Python inline on any plan, not gated behind Enterprise
- You need multiple triggers per workflow and Git version control
Choose Make if…
- You are non-technical and want a working automation today, no Docker, no JSON
- You want the widest native catalog, 3,000+ apps, with polished connectors
- Your scenarios are short (1 to 3 steps) and you value a permanent free tier
- You want managed cloud security with SOC 2 Type II and ISO 27001, no servers to patch
- You want visual branching with routers, iterators and a readable AI Reasoning panel
Frequently asked questions
Is n8n cheaper than Make?
It depends on steps per workflow. Make bills one credit per module step, so a 5-step scenario at 10,000 runs needs 50,000 credits, five times the Core base. n8n bills per execution, so the same 10,000 runs sit inside Pro at €50 a month no matter the step count. At 1 to 3 steps and modest volume, Make Core near $9 to $12 can be cheaper. The cheapest option overall is self-hosted n8n: the software is free and you pay only €5 to €15 a month for a VPS. So n8n wins for long or high-volume workflows, Make for short ones.Why is Make cheaper to start?
Two reasons. First, Make keeps a permanent free plan with 1,000 credits a month and 2 scenarios, while n8n discontinued its free cloud tier and now offers only a 14-day trial. Second, Make Core is around $9 to $12 a month for 10,000 credits, and at low step counts that stretches further than n8n Starter at €20 for 2,500 executions. The flip side shows up later: credits are consumed per step and even by polling triggers that find nothing, so the bill climbs fast as workflows get longer or run more often.What changed when Make moved from operations to credits?
On 27 August 2025 Make renamed its billing unit from operations to credits. Standard module actions still cost 1 credit each, but the important shift is AI: Make's own AI Provider modules now consume a variable amount that scales with tokens, file size and processing time, and some modules cost a fixed 2 or 10 credits. Make Code runs at 2 credits per second. Many users only noticed when the monthly bill arrived higher than the old 1:1 operations math suggested. Most comparison articles still say operations, which is why their cost estimates are stale.How does n8n per-execution billing actually work?
One full workflow run equals one execution, regardless of how many nodes it crosses. A 2-step workflow and a 200-step workflow each count as a single execution. n8n Cloud plans bundle a monthly execution quota (2,500 on Starter, 10,000 on Pro, 40,000 on Business), and Business overage runs €4,000 per additional 300,000 executions. Self-hosted Community Edition has no execution cap at all. This model rewards complex, multi-step automations: the depth of the workflow never changes the price, which is the opposite of Make's per-step credits.Can you self-host Make like n8n?
No. Make is cloud-only on its standard plans, with no self-hosted option. Its Enterprise tier mentions dedicated infrastructure, but the specific terms are not public and you would need to contact sales. n8n, by contrast, ships a free self-hosted Community Edition with unlimited executions under its Sustainable Use License, deployable via Docker, Kubernetes or a cloud VM. If full data residency or on-premise control is a hard requirement, for example in healthcare or finance, that alone points you to n8n, since Make cannot keep all data inside your own environment.Is self-hosted n8n safe in 2026?
It can be, but only if you stay patched, and 2026 proved why. CISA confirmed active real-world exploitation of critical n8n flaws in March 2026, including CVE-2025-68613 at CVSS 9.9, and more than 103,000 of around 230,000 instances were found vulnerable. Since late 2025, attackers have also abused n8n webhook URLs to deliver malware. Self-hosting puts patching, backups and OS updates entirely on you. The honest take: the Community Edition is powerful and free, but treat it as production infrastructure, not a set-and-forget toy. Make's managed cloud removes this class of risk.n8n vs Make for AI agents, which is better?
Different audiences. n8n is the deeper engine: 70+ AI nodes built on LangChain, 6 agent types (ReAct, Tools, Plan-and-Execute and more), 12 vector stores, 8 memory nodes and native MCP so your workflow can be called by Claude or GPT. That is the toolkit for RAG and complex agentic systems. Make's AI Agents are more approachable: a visual canvas with a step-by-step Reasoning panel and Module Tools that turn any module into an agent action, but no native vector stores or documented RAG layer. Build sophisticated agents on n8n, build readable no-code agents on Make.How do you migrate from Make to n8n?
There is no one-click importer, so plan a manual rebuild. Map each Make module to its n8n node equivalent, recreate triggers (remembering n8n allows several triggers per workflow where Make allows one), and move custom logic into n8n's JavaScript or Python nodes. Use the HTTP Request node for any app that lacks a native n8n node. Test with pinned sample data before going live, since n8n shows output at each node. Most teams migrate workflow by workflow rather than all at once, starting with the longest or most credit-hungry scenarios, since those gain the most from per-execution billing.n8n vs Make vs Zapier, which should you pick?
Three tiers of control. Zapier is the simplest, linear and great for non-technical users, but it gets expensive at volume. Make sits in the middle: visual branching with routers and error handlers, 3,000+ apps, no code required for most work. n8n is the most powerful and the only one you can self-host for free, with inline code and the deepest AI stack, at the cost of a steeper learning curve. Pick Zapier for dead-simple tasks, Make for visual mid-complexity automations, and n8n when you need control, high volume or advanced AI. This page compares the two heavier options head to head.Which has more integrations, n8n or Make?
Make, on raw count: 3,000+ pre-built app connectors versus roughly 1,000 to 1,200 for n8n (400+ official core nodes plus 600+ community-maintained ones). For mainstream SaaS, Make is more likely to have a polished native connector ready. n8n narrows the gap with its HTTP Request node, which turns any REST or GraphQL API into a usable node, so its practical reach is effectively unlimited even when a native node is missing. The trade-off: n8n community nodes vary in quality, and some niche Make connectors expose only part of the underlying API. Breadth to Make, flexibility to n8n.
Test both, then decide
Free to start on either side. The fastest way to know is to rebuild one real workflow on each.
Best for technical teams, long workflows, AI agents and self-hosting. Free Community Edition or a 14-day cloud trial.
Try n8n for free →Read the full n8n review →Best for no-code builders who want a working automation today and the widest app catalog. Permanent free plan, 1,000 credits a month.
Try Make for free →Read the full Make review →Affiliate links: if you sign up through them, you support our independent hands-on tests at no extra cost to you. We score both tools the same way and disclose the weak spots on each.
Get the next comparison in your inbox
Join 2,400+ makers who get our independent tool tests every week.
