A model can reason. That alone doesn’t make it an agent.
For an agent to actually do work, it needs two things:
A brain – the model doing the thinking.
And a body – the environment that lets it run, use skills, and execute inside a controlled workspace.
That’s the part people usually skip when they talk about agents.
The model handles the intelligence. The infrastructure handles the execution.
Connect the two, and you get something that can move beyond a single prompt and actually operate.
That’s the basic anatomy behind an AI agent.