The words in AI, and what they actually mean

You don't need to know any of these words to use AI well. But you'll hear them, in videos, in articles, in a coworker's sentence, and there's no reason a word should make you feel behind. Here's what they actually mean, no technical background required.

01The everyday ones

Things you'll actually click on or use yourself.

• Claude Chat. The regular way you talk to Claude. You type something, it answers. That's it, that's the whole idea underneath everything else on this page.
• Project. A home base for one piece of ongoing work. The files and instructions you give it stay put, so you're not starting from scratch every time.
• Memory. Claude remembering something you told it, so you don't have to say it again next time.
• Files. Handing Claude your actual document instead of describing it. It can read the real thing.
• Web Search. Claude looking something up right now, instead of only knowing what it already knew.
• Research. A deeper version of that. Claude digs through several sources on a bigger question and comes back with a real answer, not just one quick look.

02When it builds something beside the chat

Sometimes the answer isn't more words. It's a finished thing.

• Artifact. A separate, finished piece Claude builds next to the conversation, like a document, a chart, or a small tool, instead of just replying in text.
• Cowork. Handing Claude a whole job with several steps, like organizing a folder or putting together a report, and letting it work through the entire thing.
• Claude in Chrome. Claude looking at and working with real web pages in your browser, the same way you would.
• Claude Design. Claude creating a larger visual piece, like a poster or a page layout, not just text on a screen.

03When it's taught to do the same thing every time

• Skill. A set of instructions Claude follows the same way every time for one specific job, so nobody has to re-explain it.
• Plugin. Several of those skills, bundled together and installed at once.
• Connector. A link between Claude and an outside service you already use, like an inbox or a calendar, set up once with no technical work.

04Words you'll hear if you ever go further

You will probably never touch any of these yourself, and that's completely fine. They belong to the people building custom software with Claude underneath it. Knowing what they mean just means the word can't intimidate you anymore.

• MCP. A technical version of a connector, used by developers building custom tools. Not something a regular user ever sets up.
• Claude Code. A version of Claude built for people writing real software, working directly inside a project's actual files.
• CLAUDE.md. A note file that tells Claude Code the standing rules for one project, so it doesn't have to be reminded every time.
• Hook. A rule that fires automatically when something specific happens, without anyone asking each time.
• Agent. Claude given a goal and the room to figure out the steps itself, rather than being told exactly what to do at every point.
• Subagent. A separate copy of Claude, handed one small piece of a bigger job so the main conversation doesn't get cluttered with it.
• Agent Team. Several of those separate copies working on different parts of the same project at the same time.
• Dynamic Workflow. A way of organizing a very large job across many of those workers at once. Mostly a technical-team concern.
• API. The technical connection that lets a piece of software use Claude directly, with no person typing into a chat window at all.
• Memory Tool. The developer version of "memory" above, used when someone is building their own app that needs to remember things on Claude's behalf.
• Harness. Everything built around Claude that helps it actually do a job well. Think of it as the toolbox and the safety rails, not the model itself.