← AI Agents path

🤖 Using AI & AI Agents

Chapter 1 of 24

🤖 Chapter 1: What Are AI Agents?

Overview: agents that plan, use tools, and act

An AI agent is a system that uses an LLM not just to answer one message, but to reason over multiple steps, call tools (search, APIs, code), and use memory (conversation and sometimes long-term storage). The key difference from simple chat: the agent runs in a loop — it can decide to search, then read the result, then call another tool or finally answer the user. That loop is controlled by an agentic pattern (e.g. ReAct or plan-and-execute).

How an agent runs: the loop

User message
↓
LLM (with system prompt + history + tool list)
↓
Tool call?
→
App runs tool
→
Observation back to LLM
→(loop back up)
or
Final answer
↓
User sees reply

The agent keeps looping (think → act → observe) until it produces a final answer or hits a step limit.

Definition

Agent = LLM + tools + memory + control loop. The LLM gets the user request and (optionally) tool results; it outputs either a final answer or another tool call. The app runs tools and feeds results back until the model produces an answer or a step limit is hit.

Example: When you need an agent

Single turn: "Summarize this article" → one call, one reply. Agent: "Book me a flight to Tokyo next week" → the system might call a search tool for flights, then a booking API, then confirm with you. Multiple tool calls and turns are required; the agent coordinates them.