AI Agents

An agent is an LLM placed in a loop, given tools, and allowed to decide its own next step. Where a normal LLM call is one request and one response, an agent can plan, act, observe the result, and act again — until the task is done.

That autonomy is powerful and genuinely hard to make reliable. This section is as much about restraint — knowing when not to build an agent — as it is about building one.

The honest framing

Most “agent” problems are solved better by a fixed workflow of LLM calls. Reach for a true autonomous agent only when the steps genuinely can’t be known in advance. A predictable pipeline beats a clever loop you can’t debug.

In this section

Agent Architectures What an agent is, the core agent loop, and patterns like ReAct, planning, and reflection.

Tools & Memory Function calling, designing good tools, and the types of memory that let agents work over time.

Multi-Agent & Production Multi-agent systems, orchestration patterns, and the cost, reliability, and safety realities of shipping agents.

What you’ll be able to do

Decide whether a problem actually needs an agent, design the agent loop and its tools, and reason about the failure modes — runaway loops, cost blowups, unsafe actions — before they reach production.

Prerequisites

LLM Engineering and AI System Design.