Anthropic now has three products with "Claude" in their names.
Claude. Claude Code. Claude Cowork.
If you have used any of them, you might have wondered why they feel so different from each other, even though they all run on the same underlying model.
They share a name and brain.
However, they solve different problems, and understanding that is crucial if you are building with AI, evaluating tools for your team, or figuring out which to open.
Here is the simplest way to think about it.
Let’s dig in!
Claude: The Conversation
It is the one most people know. You open claude.ai, type a message, and Claude responds. It is a conversation. You ask a question, you get an answer.
Claude Chat interface
You paste a document, and it summarises it. You describe a problem, and it helps you think through it. Claude, the chatbot, is reactive. It waits for you.
Every interaction follows the same pattern: you send a message, Claude replies, you send another message. The context is the conversation itself.
Whatever you paste in or type is all Claude knows about your situation.
Think of it as messaging a very smart colleague on Slack. They will give you thoughtful answers, help you draft something, or explain a concept.
But they only know what you tell them. They can't access your files or run anything. They can't go off and do a task while you do something else.
For many, this is more. Need to draft an email? Summarise a report? Brainstorm ideas? Debug a concept? Claude the chatbot handles it. But there is a ceiling.
The moment your task requires Claude to actually do work in real time, such as touching your files or executing a multi-step workflow, a conversation is not enough.
This is where Claude Code comes in.
Claude Code: The Developer's Agent
Claude Code is not a chatbot. It is an agent that lives in your terminal. You install it on your machine. You point it at your codebase.
Claude Cowork interface
Then, you give it a task: "Write tests for the authentication module." "Migrate this API from Express to Fastify." "Fix the failing CI pipeline."
Claude Code does not just suggest what to do. It reads your files, writes code, runs commands, executes tests, and commits changes.
It understands your entire codebase, not just the file you uploaded, but the complete project structure, the dependencies, and the git history.
The key difference from the chatbot is autonomy.
Claude Code goes off and works. It might go through a GitHub issue, analyse the relevant code, write a fix across multiple files, run the test suite, and open a pull request.
All that from a single prompt in simple English.
You can watch it work in real time, interrupt it, or let it finish.
If Claude the chatbot is messaging a colleague on Slack, Claude Code is handing a task to a developer and letting them work in your codebase.
They have access to the tools, run things, and come back when done.
Claude Code launched as a preview in February 2025 and reached general availability in May alongside Claude 4. Within months, Anthropic reported a 5.5x increase in Claude Code revenue and a 300% expansion in active users.
But Claude Code is a developer tool. It runs in the terminal. You must know git, what a CI pipeline does, and what "run the tests" means.
(Claude has also made Claude Code available via their desktop app, which makes it very easy even for non developers to build products using Claude Code)
Also read: Claude Just Changed the Vibe Coding Game
Claude Cowork: The Knowledge Worker's Agent
Claude Cowork is what happens when you take the agent concept behind Claude Code and apply it to everyone else. Cowork uses the desktop app.
Claude Code interface
Instead of accessing your codebase, it accesses a folder on your computer.
Instead of running git commands and test suites, it reads files, creates spreadsheets, drafts reports, and organises documents. You give Cowork:
A folder of messy expense receipts, it creates a clean spreadsheet.
A collection of unstructured research notes, it produces a formatted report.
A cluttered downloads folder, it sorts, renames, and organises everything.
Like Claude Code, Cowork is an agent. It makes a plan, executes it step by step, and keeps you informed throughout the process.
Think of it as leaving a task on a colleague's desk with a note: "Can you sort through these files and pull together a summary by the end of day?"
You walk away. They do the work. They check in if something is unclear.
Cowork launched in January 2026 as a research preview for macOS. A Windows version followed in February. Since then, Anthropic has been adding connectors and plugins.
These are integrations with Google Drive, Gmail, DocuSign, and tools for specific roles, like financial analysis, HR, and design. The safety model is different from Claude Code.
Cowork runs in a more isolated environment. It cannot access anything outside the folders you explicitly grant. It asks for confirmation before taking significant actions.
Claude Code, by contrast, has broader access to your terminal and file system. That's more power and more surface area for things to go wrong.
One Brain, Three Interfaces
All three run on the same Claude model. The intelligence is identical. What changes is the interface, the level of autonomy, and the audience.
Claude gives you a conversation. You control every turn. Best for thinking, drafting, and quick tasks where you want to stay at every step. Use this when you want to think with AI.
Claude Code gives you a coding agent. You describe a task, and it executes it across your codebase. Best for developers who want to delegate real engineering work. Use this when you want AI to build.
Claude Cowork gives you a productivity agent. You point it at files and describe what you need. Best for knowledge workers who want to delegate organisational and analytical work. Use this when you want AI to organise and produce.
It is not about one being "better" than another but about how much autonomy you want to hand over, and the kind of work you are doing.
The pattern is clear. Anthropic started with a chatbot. Then they gave it to the developers. Now they are giving it in a hand for everyone else.
The connectors Anthropic announced in Feb 2026 are the beginning of Cowork becoming less of a file organiser and more of a workflow engine.
When Cowork can read your email, check a contract in DocuSign, pull data from a spreadsheet, and draft a summary in a slide deck, the boundary between "AI assistant" and "AI colleague" gets thin.