n8n + Claude Code

Module 5: n8n + Claude Code

These two tools are not competitors. They do different things well, and the boundary between them is clearer than you might expect.

What n8n Is Good At

n8n is good at anything that follows a fixed path with a clear trigger and a known output destination.

• A webhook fires, data transforms, a notification sends
• A schedule triggers, HTTP requests go out, results get stored
• An RSS feed updates, items get filtered and routed

n8n is event-driven. It waits for something to happen, then runs a defined sequence in response. The sequence is explicit — you can see every step on the canvas. It handles retries, error paths, and execution logging out of the box. It doesn’t require you to be at a keyboard.

What n8n is not good at: tasks that require extended reasoning, open-ended investigation, writing that needs to sound like you, or multi-step decisions where the path depends on intermediate results you can’t predict in advance.

What Claude Code Is Good At

Claude Code is good at tasks that require reasoning, judgement, or adaptation.

• Writing content that needs your voice
• Debugging something without a known solution
• Building a new piece of code from a description
• Investigating an unfamiliar codebase or dataset
• Making decisions based on context that’s hard to express as conditions

Claude Code is interactive. You’re present. The session has memory of what you’ve said, what you’ve built, what decisions you’ve made. It can ask clarifying questions, surface unexpected findings, and change direction mid-task.

What Claude Code is not good at: unattended scheduled work, multi-system integrations that need to run reliably at 3am, or tasks where the output needs to go to a specific destination without human review.

The Boundary in Practice

Here’s a real example from my setup.

n8n handles: The contact form pipeline. A form submission arrives, n8n validates and stores it, fires a Telegram notification. I see the notification on my phone. This runs without me being at a computer.

Claude Code handles: My response. When I sit down to reply to an enquiry, I might use Claude Code to draft something. I give it the submission context, the kind of person asking, what I think the right response is. It drafts. I edit. I send.

The boundary is attention. n8n runs when I’m not watching. Claude Code runs when I am.

Another example: the site health monitor.

n8n handles: Running the checks every six hours and sending the alert. No human involvement until something breaks.

Claude Code handles: Diagnosing the cause when an alert fires. If a site goes down unexpectedly, I might open Claude Code, share the server logs or error output, and work through what happened. That’s an investigation, not a defined sequence. n8n can’t do it. Claude Code can.

Building Systems That Use Both

The most useful pattern is: n8n as the pipe, Claude Code as the brain.

n8n moves data between systems on a schedule or in response to events. Claude Code acts on that data when it needs interpretation, creation, or repair.

Practically, this looks like:

n8n collects, Claude Code acts. n8n aggregates RSS items and sends a daily digest. I read the digest and open Claude Code to write a newsletter issue based on what caught my attention. n8n did the collection. Claude Code did the writing.

n8n alerts, Claude Code investigates. n8n fires a Telegram alert about an anomaly in traffic patterns. I take that alert into a Claude Code session, share the relevant logs, and figure out what happened. n8n detected it. Claude Code diagnosed it.

Claude Code builds, n8n runs. I use Claude Code to build a new n8n workflow — describing what I need, reviewing the exported JSON, testing the logic. Once built, the workflow runs unattended in n8n. Claude Code wrote it. n8n executes it.

When to Choose Which

A decision tree, stated simply:

Is the task triggered by an event or schedule, with a predictable output destination? Use n8n.

Does the task require reasoning, writing, or open-ended investigation? Use Claude Code.

Does the task need to run when you’re not there? Use n8n.

Does the task benefit from conversation and iteration? Use Claude Code.

Does the task involve coordinating data across multiple systems? Use n8n.

Does the task require understanding context that’s hard to express as conditions? Use Claude Code.

Most tasks fall clearly into one category. When you’re unsure, ask: could I describe every step of this as an explicit sequence with known inputs and outputs? If yes, n8n. If no, Claude Code.

One Integration Worth Building

If you’re using both tools regularly, it’s worth setting up a webhook endpoint in n8n that you can call from Claude Code scripts.

For example: a Code node in a Claude Code session generates a summary of something. Instead of copying and pasting it somewhere, the session makes an HTTP request to an n8n webhook that routes the summary to the right destination — Telegram, a database, a file on your server.

This keeps the intelligence in Claude Code (where it belongs) and the plumbing in n8n (where it’s more reliable and auditable). The webhook becomes a bridge.

What You’ve Built

By the end of this course, you have:

• n8n installed and running
• A working webhook-to-notification workflow you understand completely
• The patterns for classification, summarisation, and extraction using AI nodes
• Four real workflow blueprints you can adapt
• A mental model for when to reach for n8n versus Claude Code

The next step is building something you actually need. Start with the simplest trigger you have — a form, a schedule, a webhook — and build the minimum workflow that solves the problem. Complexity is easy to add later. Starting simple is how you learn what the tool can actually do.

---

End of course. Five modules complete.