What Claude Code is
How it differs from Claude Chat and when to use each
The core difference
Claude Chat is a conversational interface. Claude Code is an agentic coding environment — Claude can not only discuss and write code, but actually run it, read your file system, execute commands, run tests, and iterate until something works.
What Claude Code can do that Chat can’t
| Capability | Claude Chat | Claude Code |
|---|---|---|
| Write code | ✓ | ✓ |
| Execute code and see results | ✗ | ✓ |
| Read your actual files | ✗ | ✓ |
| Write files to disk | ✗ | ✓ |
| Run terminal commands | ✗ | ✓ |
| Run tests and fix failures | ✗ | ✓ |
| Use git | ✗ | ✓ |
When to use Claude Chat for code
- Explaining a concept or approach
- Quick snippets you can copy-paste immediately
- Code reviews on small amounts of pasted code
- Discussing architecture without executing anything
When to use Claude Code
- Building something in an existing codebase (it needs to see the files)
- Debugging something that requires running the code
- Multi-file changes (refactors, new features)
- Setting up a project from scratch with proper structure
- Anything involving the terminal (installs, builds, migrations)
Installing Claude Code
# Requires Node.js 18+
npm install -g @anthropic-ai/claude-code
# Run in your project directory
cd your-project
claudeClaude Code runs in your terminal and has access to everything in (and below) the directory you launch it from. It uses your Anthropic API key.
For agency work: Claude Code is highest value when working on client projects that have real codebases — the Treal Performance catalog sync, ORLA quiz system, RazorCI compliance reporting. Chat is fine for one-off scripts and explanations.