Claude

Claude Code

Anthropic's AI coding agent that works in your terminal: it reads the files in your project, proposes changes you approve, and runs the build-and-test loop with you.

Claude Code's approve-first loop You ask, Claude Code proposes a change or command and shows the diff, you approve or decline, and only then does it write or run. You ask in the terminal Proposes a change shows you the diff You approve Yes / No It writes or runs it Nothing changes on your machine until you say Yes.

Claude Code is Claude for building software, living in your terminal instead of a chat window. You start it inside a project and it reads the files it needs on demand, so there is nothing to upload. You describe an outcome ("add a login page", "find why this test fails") and it does the multi-step work: reading code, writing code, running commands, and checking the result.

The safety habit is built in. By default it works approve-first: before it changes a file or runs a command, it shows you the diff or the command and waits for your Yes. A CLAUDE.md file in the project acts as a standing brief, the conventions and context it reads every session.

It is aimed at people who work with code, but you do not have to be a senior engineer. Reach for it when the job is the code; reach for Cowork when the job is your everyday files.

Often confused with: Cowork. Claude Code is built for code in a terminal and works approve-first on a repository; Cowork works on your everyday files (docs, sheets, decks) in a folder you connect.

See it in a lesson: Claude Code

Further reading

All terms