Glossary

AI terms in plain English.

Every term a beginner meets around Claude, explained in a sentence or two. No jargon defined with more jargon.

Prompt

The message you send to an AI. A clear prompt says what you want, gives context, names the format you need, and shows an example when helpful.

Model

The underlying AI program that reads your prompt and writes the response. Claude offers a family of models; Haiku models are fastest, Sonnet models are balanced, Opus models are the most capable.

Token

The small chunk of text an AI actually processes; roughly three quarters of a word in English. Usage limits and context windows are measured in tokens.

Context window

How much a conversation can hold before the AI starts losing track of the beginning: your messages, its replies, and any files you attached, measured in tokens.

Hallucination

When an AI states something false with confidence. The practical defense is asking for sources, double-checking names and numbers, and keeping the stakes of unchecked answers low.

Artifacts

Claude’s way of putting a created thing (a document, a small app, a diagram) in its own panel beside the chat so you can see it, iterate on it, and keep it.

Projects

Folders inside Claude that keep related chats and files together, so you do not re-explain context every time. Free accounts can create up to five.

Cowork

The Claude Desktop tab where Claude works on your actual computer: organizing files, working in a browser, handling multi-step tasks, asking your permission at each step. Paid plans only.

Claude Code

Claude’s tool for structured file and build tasks. In the desktop app’s Code tab there is no terminal at all: Claude proposes changes, you review a visual diff and approve.

Connector

A permission-gated bridge that lets Claude reach something outside the chat, like your browser (Control Chrome) or your computer. You choose what to connect and can disconnect any time.

MCP (Model Context Protocol)

An open standard that lets AI assistants plug into outside tools and data sources in a consistent way. Connectors you click in Claude are often MCP servers under the hood. This site runs one at aiforlife.app/mcp.

Agent

An AI set up to take actions toward a goal, not just answer: reading pages, calling tools, completing steps. Cowork is agent-style work with you approving each step.

Permission modes

Cowork’s safety settings for how much Claude may do between approvals: Manual (approve each action), Auto (approve automatically), and Skip. Beginners start and stay on Manual.

Workflow

A repeatable sequence you and the AI complete together, like turning a folder of receipts into a summary. The whole point of this course is leaving with one that is yours.

Go further

Learn these by using them

The free lessons put every one of these terms into practice on real tasks. If you learn better alongside people, the community meets live every week.