The core of an AI agent is just a loop

Many people think agents are complicated. The core is one loop:

1. Give the model the question and current context;

1. The model decides: answer, or call a tool;

1. If it calls a tool → run it → feed the result back;

1. Repeat until it gives a final answer.

That's it. Most "frameworks" just make this loop more convenient.

The fastest way to understand it is to write it yourself — you'll find the core is a few dozen lines.

Next: giving the model hands — tool calling.