What is CLAUDE.md?

Claude Code starts every session with an empty context window, and CLAUDE.md files are one of two mechanisms that carry knowledge forward. The other is auto memory, where Claude writes its own notes. CLAUDE.md is the one a person writes, and it is meant for build commands, coding standards, project layout and other rules that should hold in every session.

Files can live at several scopes: a managed policy location for organization-wide rules, a user file in the home directory's .claude folder, a project file in the repository that is shared through version control, and a CLAUDE.local.md for personal project preferences that are kept out of git. Files in the directory tree above the working directory load at launch, and those in subdirectories load on demand. A file can pull in others with an at-sign import syntax.

The docs say Claude treats these files as context, not enforced configuration, so a rule can be missed. They suggest keeping each file under about 200 lines, writing instructions concrete enough to verify, and using a PreToolUse hook when an action must be blocked regardless of what the model decides. The docs also note that when a repository has no CLAUDE.md, Claude Code reads its AGENTS.md as the project instructions.