What Is n8n? An Agency Owner's View: n8n vs Claude
n8n is an open-source, visual, node-based workflow automation tool. It works similarly to Zapier or Make — you set a trigger (a new form submission, an incoming email, a scheduled time), and the tool automatically runs the steps that follow. The difference: you can self-host n8n on your own server, which gives you data control and cost flexibility. In my agency work, I use n8n for rule-based, repetitive tasks, and hand anything that requires judgment to an AI agent instead.
What Does n8n Actually Do?
I use n8n most in my agency work in these four areas:
- Form–CRM integration: when a web form is submitted, automatically pushing the data into a CRM, Google Sheets, or an email list.
- Notification flows: sending an instant alert to Slack or WhatsApp when ad spend on an account crosses a threshold, or when a site goes down.
- Data synchronization: keeping data in sync at regular intervals between two different systems (for example, a spreadsheet and an ad platform).
- Scheduled reports: compiling an automatic performance report every Monday morning and sending it by email.
A concrete example: when a form is submitted, n8n can be set up to push the same data to both a CRM and a Slack channel at once — the form plugin is the trigger, n8n is the bridge in between, and the CRM and Slack are the output points.
What all four have in common: they’re all repetitive and have a clear rule. Input A comes in, output B is produced, and the logic in between never changes. That’s where n8n’s real strength lies — repeating, in minutes and without errors, work that would take a human hours but requires no reasoning.
What I Hand to n8n vs What I Hand to Claude (or an AI Agent)
I go into this distinction in more detail on the n8n vs Claude Code page; here I’ll focus on n8n itself and keep it brief.
What the tasks I give n8n have in common: the logic is fixed. Which field goes where when a form is filled out never changes; a report is generated with the same query every week. I prefer n8n for this kind of work because once it’s set up, it needs no human intervention and the cost drops to almost zero — it runs as long as the server is up.
What I hand to an AI agent (mostly Claude, in my case) is work that requires interpretation, context, and judgment: assessing the tone of an ad copy, deciding how to respond to a customer email, extracting a meaningful insight from raw data. These can’t be described with a fixed flow — the input is a little different every time, and so is the right answer.
The two often overlap: n8n triggers a flow, in the middle of that flow a step (for example, summarizing or classifying a piece of text) gets sent to an AI model via API, and the result comes back into the flow. So it’s not always “n8n or the AI agent” — often it’s “AI as one step inside n8n.”
The short rule I use: if I can draw the input-output logic as a flowchart, it goes to n8n; if I can’t, it goes to an AI agent.
n8n’s Ecosystem
n8n’s node library is broad: hundreds of popular apps (Google Sheets, Slack, HubSpot, Airtable, and more) come with ready-made connector nodes. Beyond that, there are also community nodes — integrations written and shared by the community that aren’t on the official list. If you need to connect to a rare tool, it’s worth checking there first.
On the hosting side, there are two options: self-hosting (on your own server, the free open-source version) or n8n Cloud (hosted by n8n, a paid monthly plan). Self-hosting gives you data control and a cost advantage, but you have to handle server maintenance yourself; Cloud shifts that responsibility to n8n but adds a monthly bill.
The learning curve is moderate. A simple flow (form → email) can be set up in half an hour; I walked through this step by step on Your First n8n Automation. But once you move into conditional branching, error handling, or chaining multiple systems together, the visual ease runs out and logical complexity kicks in.
What to Watch Out For
n8n’s visual interface promises building automation without writing code — that holds true for simple flows. But in a complex flow (multiple conditions, error catching, converting data between APIs) you still need to be able to read JSON structure and understand at least basic programming logic. If you go in expecting to do everything without ever writing code, you’ll hit a wall at some point.
If you chose self-hosting, your automations are now tied to a server: updates, backups, and security patches are your responsibility. If the server goes down or the disk fills up, your flows stop too. This isn’t “set it and forget it” — it’s ongoing maintenance you have to own.
Finally: n8n is an automation (RPA) tool, not a decision-making mechanism. No matter how “smart-looking” the logic you add to a flow, what it actually does is follow predefined rules to the letter. Trying to force n8n to handle work that requires interpretation usually ends up in fragile, hard-to-maintain flows.
You Can Ask Me for This
This post exists purely to share what n8n is and my own experience with it — I don’t offer n8n setup or consulting as a service. If you want to see how I approach automation in general, take a look at the AI Automation Guide.
Frequently Asked Questions
Is n8n free?
The self-host version (hosted on your own server) is open-source and free. n8n Cloud comes with paid monthly plans based on usage volume.
Do I need to know how to code to use n8n?
No, for simple flows — the drag-and-drop interface is enough. But for complex flows that need conditional logic, data transformation, or error handling, basic JavaScript/JSON knowledge makes things much easier.
Should I choose n8n or Zapier?
Both do the same job; the main difference is hosting and pricing. Zapier is fully cloud-based and gets more expensive as your usage volume grows. If you self-host n8n, you can run it with unlimited operations at a fixed cost.
How do I set up my first automation in n8n?
I walked through the setup step by step on Your First n8n Automation — picking a trigger, adding the steps connected to it, and testing the flow takes a few minutes.