CLAUDE.md

CLAUDE.md is a plain Markdown file that Claude Code automatically loads as memory when it starts a session in your project. You use it to give Claude persistent, project-specific context it would otherwise have to rediscover each time — things like build and test commands, code style and conventions, architecture notes, directory layout, and “do this / don’t do that” rules. Because it is just Markdown, you edit it like any other file in your repo.

Claude Code reads CLAUDE.md from several locations and combines them. A file at the root of your project provides shared, team-wide context (commit it to version control so everyone benefits), and Claude also picks up CLAUDE.md files in subdirectories when it works in those areas. There is a user-level file at ~/.claude/CLAUDE.md for personal instructions you want across all your projects, plus an enterprise-policy location for organization-wide rules. These files layer together to form the full set of instructions Claude follows; you can also pull in other files with @path/to/file imports. See the official Claude Code docs for the exact file paths and import rules.

CLAUDE.md is part of Claude Code’s broader memory system and complements machine-readable configuration in settings.json: CLAUDE.md holds natural-language guidance for the model, while settings.json holds structured config like permissions and hooks. You can scaffold a starter file with the /init command, and quickly add a note to memory during a session by starting your message with #. Keep entries concise and high-signal — everything in CLAUDE.md consumes part of the context window.