Let's take a look under the hood and understand how Claude Code actually works. This will help you understand what's happening when you interact with Claude.

When you give Claude a prompt, it doesn't just start executing immediately. First, Claude talks to itself about the prompt, thinking through what needs to be done. Then it calls the todo manager with a list of tasks it needs to complete.

From there, Claude enters a loop. It gets a task from the todo list and works on it. For each task, Claude thinks about what needs to happen, then calls a tool - maybe running a bash command, reading a file, or searching through your code. It sees the result, thinks about what that means, and decides whether to call another tool or move on.

This continues until there are no more tasks in the todo list. Finally, Claude summarizes what it just did.

This architecture is what makes Claude Code so powerful. It's not just responding to your prompt - it's planning, executing, and adapting based on what it learns. You can actually see this process happening in the interface when Claude shows you the todo list and marks tasks as complete.