Agents · Guide
AI Agents: A Complete Guide
How AI agents observe, decide, act, and use tools — from the ReAct loop to multi-agent systems.
An AI agent is a large language model that takes actions toward a goal: it observes the task, decides what to do, calls tools or APIs, observes the result, and loops until it's done. This guide covers the agent loop (ReAct), tool use, agent memory, planning and reflection, multi-agent orchestration, and the Model Context Protocol.
Generate your own lesson →What you'll learn
- Autonomous AI Agents Explained
- The Agent Loop
- Multi-agent Orchestration
- Planning and Reflection
- Agent Memory
- AI Coding Agents: How They Work
- Browser and Computer-use Agents
- Agent Safety and Sandboxing
- Durable Agents with Workflow Engines
- Human-in-the-loop Agents
- Model Context Protocol
- Tool Use and Action Boundaries
Lessons in this guide (12)
Autonomous AI Agents Explained
An autonomous AI agent is a system that perceives its environment, decides what to do, acts, and loops — pursuing a goal without step-by-step human instruction.
The Agent Loop
Understand observe, decide, act, and update as the spine of tool-using agents.
Multi-agent Orchestration
Split work across specialists only when isolation creates real value.
Planning and Reflection
Know when planning helps agents and when it only adds latency.
Agent Memory
Separate thread state, long-term memory, and retrieved knowledge.
AI Coding Agents: How They Work
An AI coding agent is a language model wired into a reasoning-action loop that lets it plan, write, execute, and iteratively fix code to complete a programming task autonomously.
Browser and Computer-use Agents
Understand pixel, browser, and DOM action loops plus their safety limits.
Agent Safety and Sandboxing
Limit blast radius with permissions, sandboxes, approvals, and traces.
Durable Agents with Workflow Engines
Use durable execution when agent work must survive retries and crashes.
Human-in-the-loop Agents
Add approvals, review, and correction points to agent workflows.
Model Context Protocol
Use MCP to expose tools, resources, and prompts through a shared protocol.
Tool Use and Action Boundaries
Design narrow, auditable tools that agents can call safely.
Frequently asked questions
What is an AI agent?
An AI agent is an LLM wrapped in a loop that lets it take actions — call tools, read results, and decide the next step — instead of only returning text. It keeps going until the goal is met or it hands off.
How is an AI agent different from a chatbot?
A chatbot answers from the model's knowledge in one shot. An agent can DO things: query an API, run code, edit a file, then react to what happened — over multiple steps.
What is the ReAct loop?
ReAct = Reason + Act. The agent alternates a reasoning step (what to do next) with an action (a tool call), observes the result, and repeats. It's the core control loop behind most agents.