Claude Code is a
context machine.

Claude Code is Anthropic's agentic coding tool. It runs in your terminal, reads and writes files, executes shell commands, and iterates on code autonomously. It's built for engineers who want an AI that works inside their existing workflow—not a new IDE. Use it when you need multi-file refactors, automated testing loops, or codebase-wide changes that require real tool access.

01

Context is the product

Your CLAUDE.md should describe the commands, architecture, boundaries, and proof required to finish work. Keep it short enough to stay true.

CLAUDE.md
## Commands
npm run typecheck
npm run test

## Boundaries
- Preserve public APIs
- Never edit generated files

## Definition of done
- Types pass
- Critical path verified
02

The plan → build → verify loop

Start by mapping what exists. Make the smallest coherent plan. Build within explicit boundaries. Then verify the behavior, not merely the diff.

MAP→PLAN→BUILD→VERIFY
03

Leave a clean handoff

A useful completion note says what changed, what was verified, and what remains uncertain. It should help the next person move without replaying the entire session.

Stay close to the tool.

Join the first readers of the Claude Code field manual.

← Back to the workbench