Anthropic Launches Claude Cowork: Claude Code For Everyone Else

In case you missed the huge news this week, Anthropic just released Claude Cowork, and I think they nailed something important with this one.

Not because it's breakthrough new technology -- it isn’t. But because they removed the barrier that's been keeping most people from using agentic AI: the terminal.

Cowork is Claude Code (which you likely know I'm a big fan of) for general productivity, wrapped in a friendly chat interface. It works on Mac to organize folders, create documents, process data, and a lot more.

Same power, but with no command line anxiety for folks who are less technical.

So today, I want to break down:

• What exactly is Claude Cowork (and how it works)

• Why the shift from terminal matters

• What I’m excited about next

— @dharmesh

What is Claude Cowork?

In short, Claude Cowork is built on the same foundations as Claude Code and runs inside the Claude desktop app you already know.

The original Claude Code (which I use on a daily basis) is an agentic system that can be used for more than just writing code. In fact, they could have just called it Claude Agent. But, even though it can be used for doing things other than writing code, because it runs in a terminal, it's mostly developers or developer-adjacent people who use it.

Basically, Anthropic wrapped up the kernel of Claude Code and exposed it through a new feature called "Claude Cowork".

Claude Cowork runs inside of the Claude Desktop app (Mac), and does one super-simple (but powerful) thing: It gives you access to Claude Code capabilities within the friendly chat interface. Similar to Claude Code, you can let it have access to designated files/folders on your local system.

By the way, the team at Anthropic mentioned that Claude Code actually developed Cowork in about 1.5 weeks (yes, that’s AI building AI — what a time to be alive!).

Here’s how it all works (it’s simple):

1. You write a prompt

2. Point it at a folder on your computer

3. Claude can read, edit, or create files in that folder according to the given task.

The key difference from regular Claude conversations here is agency. Claude doesn't just suggest what to do -- it makes a plan and actually does the work for you.

You can queue up multiple tasks, and Claude works through them while you step away -- maybe get a coffee or take a nap. Then, it loops you in on progress and asks before taking significant actions.

It's less like chatting with AI and more like delegating to someone who works independently.

Why the shift from terminal matters

Ever since Claude Code debuted, it’s become a developer favorite. I use it pretty much daily for coding and many more tasks across my computer. (tbh they could have called it Claude Agent.)

But because Claude Code runs in a terminal, most people don't use it to its full benefit.

The terminal is intimidating. It's just a black screen with text (no visual interface or undo button), and you also need to be comfortable with command-line interfaces and file paths.

For builders like myself, that's second nature. I’ve been working in terminals for 30+ years dating all the way back to DOS. For most other people the terminal can be intimidating..

This led to a weird situation where the tool is powerful and can do tasks like organizing files or processing data, but only a small slice of people choose to touch it.

Cowork is designed to address this mismatch.

It's actually a smart move by Anthropic. They built something powerful for developers with Claude Code. Now they're making it available to everyone by changing the interface.

I don’t think it’s as big a development as MCPs (which I wrote about here), but this definitely is a major step forward towards local accessibility -- because even the best tool in the world is a fail if people don’t use it.

For me personally, Cowork doesn't change much. I'm already comfortable in the terminal and use Claude Code regularly.

But I do think that Cowork is going to be really useful for everyone else -- giving them a glimpse of what agentic AI actually feels like.

Here’s what I’m excited about: There are rumors that there’s a new version of Anthropic’s Sonnet model coming out soon. This could be a big boost for those developing agents (like me). We need even better reasoning, even better tool calling, even lower latency — and maybe even some form of “memory”. Let’s see…

I'm curious what people, especially non-developers, do with Cowork now that the terminal isn't in the way.

—Dharmesh (@dharmesh)