What is it that lets Claude Code change your code rather than only describe a change?

The principle — agency is a property of what a model is connected to, not of the model.

Why the key is correct — Anthropic puts it in one line: tools are what make Claude Code agentic, and without tools Claude can only respond with text. A tool is an action it can take, such as reading a file, editing code, running a shell command or searching the web, and each one returns a result that shapes the next decision.

Why the others are wrong — a bigger model reasons better but still cannot write to disk. The model does not write files itself; the tool layer does. And this is not a separate product, which is exactly why the distinction is worth holding.

Remember this — no tools, no actions, however good the model.

Sources — Anthropic's How Claude Code works.