What Claude can see (and what it can't)

In Lesson 2.4 you ran claude in a folder and it told you what was inside. A fair next question: if it can see that folder, what else can it see?

Claude Code can see the folder you opened it in, anything inside it, and the conversation you’re having. Nothing else, by default.

The folder you started in is called the working directory — cwd for short (the same “current working directory” you met with pwd back in Lesson 1.2). Everything Claude does revolves around it.

What it can see

Every file and subfolder under the working directory.
The conversation so far — every message, every command it ran, every result.
Files outside the working directory only if you point at them by full path. By default, it stays put.

What it cannot see

Your browser tabs, your email, your Slack, your screen. Claude doesn’t see your monitor — only your terminal.
Other folders on your computer it wasn’t pointed at.
Your previous Claude conversations. Each claude session starts fresh, with no memory of past sessions. (Cross-session memory is something you can turn on later — it’s not the default.)
Anything your account hasn’t given it access to.

Why this matters

Two reasons.

Privacy. What Claude can see is what gets sent to Anthropic to answer your question. If you don’t want a folder anywhere near Anthropic’s servers, don’t open Claude inside it.

Practicality. “Why doesn’t Claude know about that file?” almost always has the same answer: you opened it in a different folder. Move into the right folder before you start the conversation — or, when you really need to, give Claude the full path to a file you want it to look at.

The mental shortcut:

Claude lives where you opened it.

What’s next

Knowing what Claude can see is half the picture. The other half: knowing what it can do — and when it’ll stop to ask you first. That’s the next lesson.