TL;DR: Make.com wins for non-engineers building branching SaaS workflows under 10k operations/mo, where the polished UI and 2,000+ integrations earn their $9–29/mo. n8n wins for technical teams, complex multi-step logic, and any workflow where step count exceeds run count, because n8n bills per execution (one workflow run, any step count) and Make bills per operation (one step). At a 10-step workflow run 1,000 times, Make charges 10,000 ops, n8n charges 1,000 executions. Self-hosted n8n is free. The case for skipping both: if the workflow is custom logic you will run thousands of times, a 200-line worker on a $5 VPS beats both on cost and reliability.
Most "Make vs n8n" posts read like sponsored copy from one or the other. This isn't. We ship custom automations on both, plus coded workers that bypass both, and have a clear preference depending on what the workflow actually does.
Here's where each platform wins, where each one falls over, and the workflows where coding it from scratch beats both.
What each tool actually is
Make.com (formerly Integromat) is a hosted visual automation platform. Drag-and-drop scenarios, 2,000+ integrations, AI steps for OpenAI and Anthropic, pay-per-operation pricing. Designed for non-engineers building branching workflows.
n8n is an open-source automation platform you can self-host or run on n8n Cloud. JavaScript-friendly, code-first when you need it, less polished than Make for non-technical users, more powerful for complex multi-step logic.
These overlap on roughly 70% of what people build. The other 30% is where the choice matters.
Pricing in 2026
| Tier | Make.com | n8n Cloud |
|---|---|---|
| Free | 1,000 operations/mo | 5,000 executions/mo |
| Starter | $9/mo (10k ops) | $20/mo (10k executions) |
| Pro | $16/mo (10k ops, more features) | $50/mo (50k executions) |
| Mid | $29/mo (10k ops, advanced) | $120/mo (mid-tier) |
| Self-host | N/A | Free (you run the server) |
Make charges per operation, where an operation is one module run. A 10-step scenario triggered 100 times = 1,000 operations. n8n charges per execution, where an execution is one run of one workflow regardless of step count. A 10-step workflow run 100 times = 100 executions.
This is the most important pricing difference. Make's pricing scales with workflow complexity. n8n's doesn't.
Where Make.com wins
Five clear cases.
1. Non-technical operators. Make's UI is the best in the category. A non-engineer can build a branching scenario in an afternoon. n8n is friendlier than it used to be, but Make is still ahead.
2. Standard multi-tool glue. Slack-to-Notion, HubSpot-to-Google Sheets, Stripe-to-Mailchimp. Make's integrations are deeper and tested.
3. AI steps for non-engineers. Make's AI modules abstract OpenAI and Anthropic cleanly. You don't need to understand prompt structure to use them.
4. Workflows that change weekly. Make's UI rewards iteration. Drag, drop, run, debug. Faster than n8n's UI for rapid changes.
5. Smaller workflows under 10 steps. Make's per-operation pricing rewards short, simple scenarios.
Where n8n wins
The other half.
1. Long workflows. A 30-step workflow that runs 5,000 times per month is 150,000 operations on Make ($299/mo Team plan or higher). On n8n it's 5,000 executions ($50/mo Pro). The math gets ugly fast for any workflow with depth.
2. Custom code steps. n8n's "Code" node is a real JavaScript runtime. Make has a Tools section that's more limited. If your workflow needs a non-trivial transformation, n8n is the only viable option.
3. Self-hosting. n8n runs on a $10/mo VPS, on Docker, on Kubernetes, on whatever. Make is hosted only. For privacy or data residency requirements, n8n is the only option.
4. Engineer-led teams. A developer building automations professionally will outpace a non-developer on n8n by week two. The code-first paths are first-class.
5. AI workflows with retrieval, embeddings, or model choice. n8n's AI Agent node, vector store nodes, and model flexibility (OpenAI, Anthropic, Cohere, Mistral, local models) are best-in-class. Make's are catching up.
Where both start to fail
Three patterns where you should skip both.
1. Workflows you'll run 100k+ times a month. At that volume, even n8n self-hosted starts to need real infrastructure attention. A coded worker on Vercel or a small VPS is cheaper, faster, and more debuggable.
2. Workflows that touch sensitive data. Anything healthcare, financial, or regulated. The audit, compliance, and data residency story for both platforms is improving but not where a custom worker with proper logging and access control is.
3. Workflows that are core product, not glue. If the automation IS the product (a customer-facing AI workflow, an inference pipeline, a real-time data system), don't build it on a no-code automation platform. Build it as a service.
Cost example: a real lead-routing workflow
A typical workflow we ship: form fill → enrich the lead via Clearbit → score against ICP rules → route to a Slack channel → write to HubSpot → trigger a templated outbound email.
Six steps. Triggered ~500 times a month for a small B2B service business.
| Platform | Monthly cost | Maintenance |
|---|---|---|
| Make.com Pro | ~$16 (3,000 ops) | Self-serve |
| n8n Cloud Starter | ~$20 | Self-serve |
| n8n self-hosted | $10 (VPS) | You operate it |
| Coded worker | $20 (Vercel) + LLM API | One-time build, light maintenance |
At 500 runs/mo, all four are cheap. At 50,000 runs/mo, Make is $599+ (Team or Enterprise), n8n Cloud is $50–$120, n8n self-hosted is still $10, and a coded worker is $20–$40 plus LLM cost. The gap is structural.
What we recommend at each stage
| Stage | Default |
|---|---|
| 1 person, occasional automation, mostly glue | Make.com |
| 2–10 person team, multiple workflows, ops mindset | n8n Cloud |
| Engineer-led team or compliance-sensitive | n8n self-hosted |
| Workflow is core product or runs at high volume | Coded worker |
When we'd ship a coded automation instead
We've covered this in Zapier vs custom AI automation, and the same logic applies to Make and n8n. The break-even shifts depending on volume, LLM intensity, and the complexity of the failure modes you need to handle.
Short version: under 5,000 runs/month with simple logic, no-code is the right call. Above that with LLM steps and meaningful failure handling, custom usually wins.
Three questions to decide this week
- How long is the longest workflow, and how many times will it run? Long and high-volume: n8n. Short and lower-volume: Make.
- Is anyone non-technical going to maintain this? If yes, Make. If no, n8n.
- Do you need self-hosting? If yes: n8n self-hosted. If no, n8n Cloud or Make.
If you're hitting the ceiling of either platform and want to know whether a coded automation is the right next step, send us the workflow and the run volume. We'll tell you within two days whether the math works.