n8n

If you've built automations in Zapier or Make, you know the moment it stops working: the visual builder can't express your logic, and the only way out is a separate script running somewhere else. n8n's pitch is that you shouldn't have to pick a side. Its homepage says it bluntly: "Other tools limit you to either a visual building experience, or code." In n8n you build on a canvas, but you can write JavaScript or Python anywhere in a workflow. And unlike most hosted automation tools, you can run the whole thing yourself. The source is on GitHub, it deploys with Docker, and it'll run on-prem or fully air-gapped if your security team insists.

The second big difference is how it handles AI. n8n positions itself as a platform for AI agents "you can see and control." Every step of an agent's reasoning shows up on the canvas, you can inspect the prompt, response and outcome of every execution, and you can add human-in-the-loop approvals before an agent does something risky. It supports multi-agent setups and RAG, lets you swap cloud models for offline ones without rebuilding, and connects to anything that speaks MCP. That's a more serious agent story than "add an AI step to your Zap." The social proof is solid too: 207k GitHub stars, a 4.7/5 rating on G2, and case studies from Huel (1,000 hours of manual work saved) and Vodafone (£2.2 million saved in threat intelligence).

The trade-off is that n8n is built for technical teams, and it shows. You'll get the most out of it if you're comfortable reading execution logs, writing the occasional code node, and thinking about where your workflows run. If you are, it's one of the few automation tools that won't box you in as your workflows get more complicated. If you just want to connect a form to a spreadsheet in five minutes, it's probably more tool than you need.