AI Agent

An AI agent is a system that uses an LLM as the planner inside a loop. The LLM decides what to do next, calls tools (APIs, functions, search), observes results, and continues until the task is done — without explicit step-by-step programming.

The simplest agent loop: (1) **plan** — LLM decides next action, (2) **act** — call the chosen tool, (3) **observe** — return result to LLM, (4) **repeat** until done or stuck. This is sometimes called "ReAct" (Reason + Act).

Multi-agent systems coordinate several agents — e.g., a "researcher" agent gathers data, a "writer" agent produces a report, a "reviewer" agent critiques it. Frameworks like LangGraph make this manageable in production.