N8n – Flexible AI workflow automation for technical teams

n8n has been good but not great in our organization (and we pronounce it "Nathan", to answer someone else's question). It's effectively a self-hosted version of Zapier and has quite a few built-in integrations. It's a bit more annoying to use than Zapier (but the price is right), and the AI features are currently about like the AI features of every other product: basically sufficient to tell investors "we do AI!" but not anything you're going to actually use.

The one frustration we have with n8n is trying to create custom "apps" (triggers or destinations for workflows). It's clear that the custom apps are an afterthought and have gone through multiple iterations of "here's the best way to do it", and you end up having to just keep trying until you get it to do what you want. Annoyingly, there's no way to manage custom apps in the interface itself - you have to create a Javascript module and then inject it into a .npm directory somewhere inside of the applications Docker container, which just doesn't feel very "professional".

If n8n would add some kind of admin interface for managing custom apps - especially just supporting basic use cases like specifying a REST API as a reusable custom app - it would be great, but still has a ways to go in terms of features (like better user permissions management as part of the lackluster SSO) before it's truly going to be an enterprise grade solution.

That said, we tried Windmill first and while it was cool for the devs who were able to see the vision, the non-technical users hated it and have heavily praised n8n once we created a custom app to let them integrate with our system.

Overall I would say n8n is worth trying if you need something like this, but expect to do some tinkering if you go beyond what it does out of the box.

Before it was moved to GitHub Actions, we had a multiple-step deployment flow created in Node-Red, I believe you can achieve more or less the same thing with n8n. Never tried tho, it wasn't necessary anyway. If I really really simplify it's like, checkout, run some tests, deploy to a preview env, run some more tests, report back to slack.

On a personal level, I use it to automate booking a tennis court. It logs in, selects the date and time (pre-defined), adds a partner/opponent, books the court, pings a service that generates a calendar entry. If I decide to cancel the booking, it again pings the service and removes the calendar entry. I needed to bend some "nodes" to do what I need within the same workflow (such as create event runs once a day but update event runs in every 10 minutes).

In the past, I also used it to detect service interruptions with my ISP. I don't use that ISP anymore, so this workflow is redundant, but it was checking an API and if there's an entry, sending it as a push notification (via ntfy) to my phone.

Looks like a product you'd end up with if a few years back you thought to yourself "how do I make a business combining AI and pipelines?". I don't hate AI as such, I just don't love how it has to be shoved into every product or tech imaginable these days.

This sounds like excellent evidence that they picked a terrible name and should change it.

Wow. This project was the cause of a very long and intense discussion about mis-use of the term "open source". See https://github.com/n8n-io/n8n/issues/40#issuecomment-5397146... for details (lands mid-thread -- you might want to scroll back to see the start, and if you read the whole thing to the end then you deserve some sort of award!).

TL;DR: The author originally tried to call n8n "open source" but while using a non-open-source license. After much discussion, he kept the license but stopped using the label "open source", to the relief of many people.

That half-decade-old thread is still what I point to when I want to explain to someone why preserving the specificity of the term "open source" matters.