If you’ve downloaded the Claude desktop app, you’ve probably noticed three tabs at the top: Chat, Cowork, and Code.
Most people live in Chat and never touch the other two.
That’s leaving a lot on the table.
Here’s the one-sentence version of each — and when to use them.
Chat is for thinking.
No file access, no autonomy. Just a conversation. Brainstorming, drafting, researching, working through a decision — this is your mode. If you want to think out loud with something that pushes back intelligently, Chat is it. Fast, frictionless, no setup.
This is where I spend most of my time.
Cowork is for doing.
Claude Chat is conversational. Claude Cowork is operational. Chat responds to prompts. Cowork delivers outcomes.
Before taking action, Claude generates a step-by-step plan showing how it intends to complete the work. You can approve the plan, adjust it, or cancel it entirely. So you stay in control — it’s just doing the actual work instead of describing it to you.
Good for: organizing files, building reports, automating the repetitive stuff across your desktop apps. It runs in a virtual machine that’s isolated and protected — safer than giving an AI direct access to your machine.
Code is for builders.
Claude Code is Claude as an agent that runs directly on your computer. You type your request in plain text, and Claude reads files, writes files, runs commands, and can take over full workflows on your machine.
If you’re not a developer, you probably don’t need this yet. But if you’re building products, this is where the real power lives. It runs your tests, your build, your linter. It sees the errors and fixes them. The feedback loop is tight.
The mental model I use:
- Chat = thinking partner
- Cowork = virtual assistant who can actually do things
- Code = autonomous engineer on your codebase
Same AI. Three completely different modes. Pick the right one for the job.
— Warren