CLAUDE.md: Your Project's Memory

A CLAUDE.md file is project memory the agent reads on every turn. It is where you write down the things you would otherwise repeat in every session: how to run tests, what conventions to follow, which directories are off limits.

What belongs in it

• Build and test commands the agent should use.
• Code style and architecture conventions specific to this repo.
• Known gotchas ("the API client retries, do not add your own").
• What not to touch (generated files, vendored code).

# Project: payments-api

## Commands
- Test: `npm test`
- Type check: `npm run typecheck`

## Conventions
- No `any`. Prefer discriminated unions.
- All money is integer cents, never floats.

## Do not touch
- `src/generated/**` is codegen output.

The same idea works under the open AGENTS.md name, which many other agents read too. Pick whichever your tools support and keep it lean.