Definition
An AI agent combines a model with instructions, a set of tools (APIs, databases, a browser, other agents) and a loop. On each turn the model decides whether to call a tool, reads the result and decides again, until it reaches the goal or a stopping rule.
What makes something an agent is that the model controls the sequence of steps. If the steps are fixed in code, it is usually better described as an agentic workflow. Agents suit open-ended tasks where the path cannot be predicted in advance, at the cost of higher latency, higher spend and more ways to go wrong.