A teammate says Claude Code is simply Claude under a different name. Which part of the setup are they leaving out?

The principle — Claude is the model; Claude Code is what has been built around it.

Why the key is correct — Anthropic names the surrounding layer the agentic harness and defines it as the tools, context management and execution environment that turn a language model into a capable coding agent, adding that Claude Code is the harness and Claude is the model inside it. That is why the same model behaves so differently in a chat window: the model has not changed, its surroundings have. The harness is what supplies file access, shell execution, permission gating and the loop that chains actions together.

Why the others are wrong — treating them as identical cannot explain why one can edit your repository and the other cannot. There is no separate coding model; the weights are the same. And a larger window changes how much can be held, not whether anything can be done, which is a capacity question rather than a capability one.

Remember this — when something can be done in one place and not the other, the harness is usually the reason.

Sources — Anthropic's Claude Code glossary.