Can Claude Read a Folder of Files?

Lines of source code displayed on a laptop screen

I get this question a lot, and it’s a fair one to be confused about — “Claude” means different things in different contexts, and each one has a completely different relationship with your filesystem. Here’s an honest capability map so you know exactly what to reach for.

claude.ai (the web app): no local folder access

If you’re using Claude through your browser at claude.ai, it cannot read a folder on your computer. Full stop. There’s no “connect a folder” button, no watched directory, no mount point. What you can do is upload individual files into a conversation or add knowledge files to a Project — but you’re uploading them manually, one at a time (or as a ZIP).

The web app also can’t watch your filesystem. If you add a file to a folder on your Mac or PC, Claude has no idea unless you re-upload it. This trips people up because it sounds like such an obvious feature — “just let me point Claude at my /Documents/project folder” — but it’s a hard boundary of the browser-based product.

One nuance worth knowing: you can add files up to 30 MB each to a Project’s knowledge base, and they persist across every chat in that project. That’s useful for reference docs, design specs, or code snippets. But it’s still manual — not a live folder connection.

For a deeper dive on how to work with files and context in Claude’s web app, see how to add folders to Claude.

Claude Code: real filesystem access

Claude Code is a different product entirely. It’s a CLI tool you install on your machine and run from a terminal. When you start a session, it operates in your current working directory and has genuine filesystem access — it can read files, write and edit them, rename them, create new ones, delete them.

If you point it at a project folder and say “read all the TypeScript files in src/ and summarize the architecture,” it actually does that — no uploads, no manual selection. It walks the directory tree itself. I use it daily for exactly this kind of task: bulk renaming files to a consistent convention, refactoring across a whole folder, reading config files and proposing edits.

What it can’t do is monitor a folder in the background. Claude Code is interactive — you ask it to do something, it does it, and that’s the session. It doesn’t sit idle watching for new files to appear. If you need that kind of trigger, you’d write a small shell script or file watcher that calls Claude Code when something changes.

The sibling article adding a folder of files to Claude has concrete workflows for getting a whole directory’s worth of context into a session.

Claude Desktop + MCP connectors

Claude Desktop (the native macOS and Windows app) sits between the web app and Claude Code in terms of capability. On its own it has the same restrictions as the web app. But it supports MCP (Model Context Protocol) connectors — plugins that give Claude access to local tools and data sources, including the filesystem.

With the right MCP connector configured, Claude Desktop can read from a local directory. Setup requires some technical comfort — you’re editing a JSON config file and installing a connector — but it works. Like Claude Code, it reads on demand rather than watching passively.

Organizing your claude.ai chats is a different problem

One thing I want to be clear about: NorthLab Folders — the browser extension I built — is not for connecting local folders to Claude. It’s for organizing your chats inside claude.ai and chatgpt.com. The two problems sound similar but are completely different.

If what you need is a way to group and find your conversations — not local files — NorthLab Folders adds a real folder sidebar with subfolders, pinned chats, and date stamps. Your conversations stay in your browser; nothing leaves your machine.

Which should you use?

Here’s the quick decision tree:

  • Want Claude to read, edit, or rename files on your computer? → Claude Code (CLI) or Claude Desktop with an MCP filesystem connector.
  • Want to give Claude reference documents that persist across web chats? → Upload to a Project knowledge base on claude.ai.
  • Want to organize your claude.ai conversations into folders? → NorthLab Folders extension.
  • Want to monitor a folder for new files automatically? → Nothing native; you’d script it around Claude Code.

The frustrating-but-honest answer is that “can Claude read folders” has three different answers depending on which Claude you mean. The web app is sandboxed by design. Claude Code and Desktop+MCP are where real filesystem work happens.

Frequently asked questions

Can Claude access files on my computer?

The claude.ai web app cannot access files on your computer — you have to upload them manually, one at a time. Claude Code (the CLI tool) does have real filesystem access and can read, write, and rename files in any directory you give it. Claude Desktop with MCP connectors can also connect to local folders.

Can Claude edit or rename files in a folder?

Claude Code can edit, rename, create, and delete files in a project directory. You point it at a folder and it works directly on the real filesystem. The claude.ai web app cannot do any of this — it only sees what you upload to it.

Can Claude watch a folder for new files?

No version of Claude has native folder-watching built in. Claude Code can read a directory’s current state when you ask it to, but it doesn’t sit in the background monitoring for changes. You’d need to script that trigger yourself and pipe it into Claude Code.

Does claude.ai read local folders?

No. The claude.ai web app has no access to your local filesystem. You can upload individual files or a ZIP archive to a Project’s knowledge base, but Claude cannot browse, watch, or mount a folder on your machine through the web interface.