n8n, make.com, and claude: what actually matters in 2026

People keep asking me which one wins: n8n, Make, or Claude.

That’s the wrong question.

n8n and Make are workflow automation platforms. You build something once, it runs on schedule, handles errors, moves data between systems, and keeps going without you touching it. Claude is a different layer. It’s an AI product suite for knowledge work and coding. It can handle one-off analysis, work across files and apps, and help build the workflows themselves.

They solve different problems. Most agencies doing automation seriously end up using all three.

What each one is for

n8n is the developer pick. It’s self-hostable, has 500+ integrations, and gives you code nodes when the visual builder runs out of room. If a client wants tighter control over data, infrastructure, or logic, this is usually where you go.

Make is the visual one. Flowchart canvas, routers, iterators. A client can open a Make scenario and usually follow what it does. That matters more than most engineers want to admit.

Claude is the AI layer. Claude Cowork can work across files and apps on your computer. Claude Code works in the terminal, IDE, desktop app, and browser for anything that touches a codebase. It’s not the same kind of platform as n8n or Make. It’s closer to an operator that can do the messy thinking, drafting, analysis, and first-pass build work those platforms still need.

How they fit together

Claude handles the unstructured layer that n8n and Make don’t handle by themselves: a folder of PDFs, a pile of call transcripts, a Notion workspace nobody has cleaned up, a set of contracts that need comparison before anything can be automated.

n8n and Make handle the repeatable system layer: triggers, branching logic, retries, routing, syncing, notifications, and the things that need to happen every time, not just once.

And both n8n and Make can call Claude inside a workflow. So the stack composes cleanly.

How an automation agency uses all three

I’m building one right now, so this isn’t theoretical.

Sales calls go through Claude first. It reads the transcript, maps the prospect’s process, and drafts a proposal with scope and ROI assumptions. What used to take two hours now takes closer to fifteen minutes with review.

Make handles a lot of the client-facing automation because it’s readable. Lead capture, follow-up, internal alerts, CRM updates, handoffs.

n8n comes in when a client wants more control over hosting, governance, or logic than a pure cloud-first visual tool gives them.

The tooling cost is still low relative to what you can charge. Make plans can start cheap. n8n’s community self-hosted option is free. Claude is still inexpensive compared to billable delivery time. So the economics work.

I’m charging $2,000 to $5,000 to build, plus a retainer to maintain it.

The margin is in the builds.

Give Claude a few strong examples and it can draft real workflows, including JSON you can often adapt directly. You still review it. You still test it. You still fix edge cases. But you’re not starting from a blank screen.

The split that’s been working for me is simple: most builds start in Make because clients can read them, and a smaller share goes into n8n when the client needs self-hosting or the logic gets ugly. Claude writes the first draft of a lot of it.

Where each one breaks

Claude is not the same thing as an always-on workflow engine. It can help build, analyze, schedule, and assist, but it’s not your replacement for a proper automation backbone.

n8n does not magically understand the messy context sitting in files, docs, and human conversations unless you wire that in.

Make is strong visually, but it’s primarily a cloud-first product, which matters when a client wants infrastructure they control more directly.

Each one is weak where the others are strong. That’s why the stack works.