What is Claude Code, explained to someone who has never opened a terminal

Claude Code is Anthropic's tool for working on a task that begins with files. You choose a working folder. You describe the result you want. Claude can read the files in that job and make a new file for you to inspect.

The folder is a useful first boundary, not the whole security model. In the practice run below, I chose one local folder of fictional files. The request also said which files had to stay unchanged and which outward actions Claude could not take.

If you prefer a graphical route, start with Anthropic's Claude Code Desktop app. Claude Code 101 expects basic familiarity with a code editor and the command line, so learn those before you use that course.

On 11 September 2026, I opened a local practice folder in Claude Code Desktop. The folder held six fictional training enquiries and one response guide. No real customer information was involved.

I asked for one action list. The request named the output file, asked for every message to appear once, and said that missing information had to stay visible. It also said not to contact anyone, change a booking, use a connected account, or schedule a reply.

Claude Code listed the inputs and wrote outputs/action-list.md. The report placed the urgent booking and the accessibility question first. It left seat availability, captions, and a quiet room as questions for a person instead of making them up.

Start with a job you have already completed by hand. Copy the source files into one folder. Ask for a draft, a list, or a summary in a named output file.

Say what Claude may read. Say what the result should contain. Say what must stay uncertain. Then say what it may not do. A first request can ask for a report and forbid sending, booking, charging, deleting, or using another account.

A basic chat conversation begins with the text you put into the chat. You bring the question, add the details you want to share, and choose the next message.

Claude Code can begin with a folder you selected for a specific job. In the recorded practice run, it read the supplied files and wrote a report beside them. This is the local workflow I tested here.

The report was useful because the original enquiries and response guide remained beside it. A polished answer alone cannot tell you whether the job used the right facts.

Anthropic's current setup guide documents the Desktop app and the command-line installation routes. The same guide says that Claude Code access requires a Pro, Max, Team, Enterprise, or Console account. The free Claude.ai plan does not include Claude Code access.

The recorded Desktop example showed a Code workspace, a chosen local folder, activity for reading inputs and writing the output, and the action list. It shows one real run from 11 September 2026. It does not promise that your screen will look the same.

Read every permission request. In a later practice routine, Claude Code asked for one-time approval before it edited an existing generated report. I approved that edit, checked the result, and paused the routine. A saved routine is not proof that a job can run forever without a person.

Start with Anthropic's Claude 101. Its prerequisites allow a Claude account on any plan, including Free, and no previous experience with Claude or another AI assistant. The course page says some hands-on practice needs a paid plan.

Claude 101 gives a nontechnical reader a first conversation and a prompt before the more technical route. Anthropic's Claude Code 101 is for someone who already has basic familiarity with a code editor and the command line. Its hands-on route requires a Pro, Max, or Enterprise account, or an API key.

Use the first lesson after you have a small folder job in mind. Then repeat that job with copies of files you already understand. My Claude Code course is for the next stage, where you want more practice with permissions, checks, and repeatable systems.