Building a Simple AI Agent

People overcomplicate AI agents. They’re actually stupid simple. Here’s the entire

thing in about a minute. An AI agent is just three parts: A brain — the model.

A plan — the reasoning layer. And tools — the stuff it can actually use.

You give it input → it thinks → it acts. That’s it. Let’s build one

in n8n fast. Drop an AI Agent node. It errors — no brain yet.

Add a Chat Model, connect your API key, done. Say “hello.” Works.

Say “My name is Matt, help me send an email.” Then ask “What’s

my name?” It blanks — because there’s no memory. Add Window Buffer Memory.