Agent: Definition and Examples

An agent in artificial intelligence refers to a system capable of perceiving its environment, making decisions, and executing actions autonomously to achieve a defined goal. Unlike a simple language model that merely generates text in response to a query, an agent can break down a complex task into sub-steps, choose appropriate tools, and iterate until the desired result is obtained.

The concept of an agent is based on a fundamental loop: observe, think, act. The agent first analyzes the situation (the prompt, context, previous results), devises an action plan, then executes that plan using various capabilities such as web search, code execution, file reading, or calling external APIs. At each step, it evaluates the outcome and adjusts its strategy if necessary.

In prompt engineering, understanding how agents work is essential because it radically changes how one interacts with AI. Instead of writing a single prompt and hoping for a perfect response, you can entrust an agent with an overarching mission and let it orchestrate the necessary steps itself. This involves formulating clear objectives, defining constraints, and providing relevant context rather than step-by-step instructions.

Agents represent a major evolution of generative AI. Tools like Claude Code, AutoGPT, or custom ChatGPT assistants illustrate this trend. They enable the automation of entire workflows — from software development to document research — by combining reasoning, memory, and tool use in a continuous execution loop.