memory-design

Memory design establishes a 4-tier hierarchy: enterprise memory loads first, followed by project memory, project-specific rules, user memory, and project-local memory. This loading order ensures scalable organization where instructions cascade from broad governance to task-specific guidance, with each tier overriding prior levels when needed.

CLAUDE.md holds project-wide context and workflows; .claude/rules/ organizes path-specific rules that apply conditionally based on file location. Use CLAUDE.md for general project knowledge and next_step decisions, and .claude/rules/ for directory-scoped constraints that Claude Code discovers recursively during memory loading.

Enterprise memory applies organization-wide; project memory is repository-scoped; project rules are path-specific; user memory is individual-level; project-local memory is session-scoped. Each tier narrows scope and increases specificity, enabling progressive disclosure where Claude Code loads only relevant memory for the current workflow.

Path-specific rules use recursive discovery: Claude Code scans .claude/rules/ directories up the file hierarchy and loads applicable rules based on your current working path. This mechanism supports modular memory organization, ensuring project-wide rules apply globally while subdirectory rules override them for localized workflows.

Yes. Memory design supports enterprise-level organization through its 4-tier hierarchy: enterprise memory sets governance for all teams, project memory isolates repository-level workflows, and path-specific rules let teams within the same project apply localized constraints without conflicts.