memory-hierarchy

Claude Code's memory hierarchy prioritizes files by location: Enterprise policy, Project memory, Project rules, User memory, User rules, and Project-local memories. Organize by creating CLAUDE.md files at each level and storing rules in `.claude/rules/` directories. This ensures context loads in the correct priority order for your team and monorepo workflows.

Store rules in `.claude/rules/` and use YAML frontmatter with glob patterns to scope them to specific file types or directories. For example, add a rule targeting `src/api/**/*.ts` files to enforce input validation. Rules apply automatically based on the files you're working with.

Use the `@path/to/file` import syntax to link documentation and other memory files. Supports relative, absolute, and home-directory paths. Import statements enable modular organization and cross-reference of knowledge across Enterprise, Project, and User memory locations.

Yes. Claude Code distinguishes between gitignored and non-gitignored memory files. Store sensitive or local-only context in gitignored memory files within your project structure, while keeping shared team memory in version-controlled CLAUDE.md and rules files.

Claude Code automatically loads CLAUDE.md and rules files through upward recursive discovery. It searches from your current project directory up the directory tree, loading memory files and rules at each level according to the memory hierarchy priority order.

Memory hierarchy enforces strict precedence: Enterprise policy overrides Project memory, which overrides User memory, which overrides Project-local memories. Explicit memory-priority ordering ensures that higher-level policies and rules always take effect, enabling consistent enforcement across team and individual workflows.