How AI agents work

AI explained

An AI agent takes a goal, thinks, uses tools and checks what happened, again and again, until the job is done. Watch one work, then pick the way in that suits you.

I am new to thisThe kitchen storyA restaurant kitchen builds itself on one screen while a chef shows what an agent does. Everyday words only.
I want the full pictureThe technical deep diveEight chapters in precise terms, with dated facts and sources. About 9 minutes, searchable.

Try it

Give an agent a goal

How the loop works, in chapter 4

Pull a term from the glossary

Every term has a drawing, a definition and references. Here is one at random.

Agents and loops

Agent loop

The cycle that turns a model into an agent: look, act, see the result, repeat.

Read the whole term

Popular starting points

MCP
Agent loop
Prompt injection
RAG
Human in the loop
Guardrails

Browse the whole glossary

Then go deeper

Long guides, each one question from start to finish: how agents work, what fills their context, how to build, secure and test them.

How an agent works23 min read
Context engineering22 min read
Build your own24 min read
Securing agents25 min read
Testing and running25 min read
Easily confused21 min read
Where to learn more15 min read
All the guidesOpen the shelf