Five AI Words in Plain English

You do not need to memorize technical definitions. Think about what each part does in a real task.

Open the full-size image if you want to zoom in or save it.

Five words, with examples

1. Harness: the workspace that gives an AI model the files and tools it needs to do a job. Codex is one. If you ask Codex to fix a bug, the model reasons through the change while Codex lets it read your project, edit files, and run checks. VS Code is an editor you may use beside it.
2. API: a way for one app to ask another app for something. A weather app asks a weather service for today's forecast and receives the result. You do not have to open the weather service yourself.
3. CLI: a tool you control by typing a command instead of clicking buttons. In a terminal, you might type ls to see the files in a folder.
4. MCP: a common way an AI app connects to outside tools. Connect a calendar through MCP, and the AI app can use the calendar tools you allowed. Check what access you grant before connecting it.
5. Skill: saved instructions for one repeatable job. A website-checking skill might tell an AI tool to open the page, test the buttons, check it on a phone, and report what broke.

How they fit together

The model does the thinking. The harness gives it a place to work. A skill tells it how to handle a particular job. MCP can connect it to another tool. Apps may use APIs to exchange information. A CLI is one way you can give a tool commands. A simple task may use only one or two of these.

If you remember one thing, make it this: the model does the thinking. The other pieces help it do the job.