Most AI interactions follow a simple pattern: you send a message, the model generates a response, the exchange ends. The model has no memory of what came before, no awareness of what happens next, and no ability to do anything in the world beyond producing text. This is useful. It is also a ceiling.
An AI Agent breaks through that ceiling by adding something the conversational model lacks: the ability to act. An agent doesn't just generate a response to your input. It perceives its environment, decides what to do, takes an action, observes the result, and decides what to do next. It can run a search, write and execute code, send a message, update a database, or call an API. It can do these things in sequence, adjusting its plan based on what it finds. It is pursuing a goal, not answering a question.
The distinction matters because it changes what AI can actually do. A conversational model can tell you how to book a flight. An Autonomous Agent can book it. A conversational model can describe the steps to analyze a dataset. An LLM Agent can write the code, run it, interpret the output, and iterate until the analysis is complete. The gap between describing a task and completing it is the gap between a conversational model and an agent.
This capability compounds when agents work together. Multi-Agent AI systems coordinate multiple specialized agents, each handling a part of a larger task. One agent might handle research, another drafting, another review. The coordination between them allows the system to tackle problems that would be too complex or too long for any single agent to handle alone.
The articles in this section cover what agents are, how they are structured, and what the move from single-turn AI to autonomous AI actually means. The sections that follow cover how agents make decisions and how they connect to the tools and systems that let them act in the world.
