n8n
Fair-code workflow automation platform that connects your apps and services. Free to self-host with no per-execution costs, and surprisingly powerful for AI-augmented workflows.
My take
n8n is the backbone of our automation stack. We generate hundreds of contracts and invoices per week through it, each in about 45 seconds. The self-hosted model means we pay nothing per execution, which matters when you're processing thousands of automations monthly. If your team has even a little technical ability, n8n beats Zapier and Make on both cost and flexibility.
At a Glance
Pros
- + Self-hosted option means no per-execution fees and full data control
- + Visual workflow builder makes complex automations accessible
- + 500+ native integrations including AI models, databases, and business tools
- + AI nodes let you mix deterministic automation with LLM-powered steps
- + Active community and rapid development, with source-available code you can inspect
Cons
- − Self-hosting requires some technical setup and maintenance
- − Visual builder can get unwieldy for very complex workflows
- − Cloud is metered by executions, so the cost adds up with many workflows
- − Learning curve steeper than Zapier for non-technical users
Best for: Teams that need reliable, repeatable automations without per-execution costs when self-hosted — especially when combining deterministic processes with AI
Full review
n8n sits in a category that most people don’t think about carefully enough: workflow automation. Not AI, not custom code — the middle layer where repeatable, deterministic processes live. And it’s the layer that often delivers the most immediate, measurable ROI.
Why n8n over Zapier or Make
The short answer: self-hosting and cost structure.
Zapier and Make charge per execution. That’s fine when you’re running 50 automations a month. It gets painful fast when you’re running thousands. n8n’s self-hosted version has zero per-execution cost. You pay for the server (which is cheap) and that’s it. n8n Cloud meters executions too, so that advantage applies to self-hosting only.
The longer answer: n8n gives you more control. You can build complex branching logic, error handling, and retry mechanisms that would be awkward or impossible in simpler tools. And because n8n calls itself “fair-code” (source-available under its Sustainable Use License, free to self-host), you can inspect exactly what it’s doing — no black boxes. It isn’t open source in the strict sense, and the free Community Edition leaves out some enterprise features such as SSO and Git version control.
Where n8n fits in an AI stack
Here’s something I had to learn the hard way: not everything needs AI. When I first got excited about LLMs, I wanted to throw AI at every problem. But for generating a contract from structured data? For sending a payslip based on a calculation? For triggering an alert when a database value changes? AI is overkill. These are deterministic processes. They should run the same way every time, reliably.
n8n handles this layer. At our company, it processes hundreds of employment contracts per week, generates invoices, handles electronic signatures, and runs data quality checks. Each contract takes about 45 seconds to generate. No AI involved — just reliable automation.
But n8n also has AI nodes. So when you do need judgment — summarizing a document, classifying a support ticket, extracting information from unstructured text — you can mix AI steps into an otherwise deterministic workflow. Best of both worlds.
The framework: code vs. automation vs. AI
I think about it in three layers:
Code (your application) — for core business logic. Matchmaking algorithms, pricing calculations, platform features. Things where you need a clear master system, full determinism, and full control.
Automation (n8n) — for repeatable processes with deterministic outputs. Contracts, invoices, signatures, alerts, data pipelines. Needs to run reliably, every time, the same way.
AI (Claude, Gemini, etc.) — for when you need judgment, language understanding, or dealing with ambiguity. Customer service, screening, predictions, content generation.
“AI for everything” is as wrong as “AI for nothing.” The discipline is in picking the right tool for each job. I wrote a full guide on this framework.
Who should use n8n
If you’re a solo practitioner with simple automations, Zapier is probably easier to set up. But if you’re a team running dozens of workflows with real volume, n8n’s cost model and flexibility make it the better choice. The sweet spot is teams with at least one person comfortable with basic technical concepts — you don’t need to be a developer, but “connecting APIs” shouldn’t scare you.
Getting started
The fastest path: try n8n Cloud (there’s a free trial, no free tier) to learn the interface. If you like it and your volume justifies it, move to self-hosted. The community templates give you a head start on common workflows.
Added: 2026-03-17 · Last updated: 2026-10-22