memory-and-rules

Claude Code Memory and Rules Reference

Claude Code has two kinds of persistent memory:

• Auto memory — Claude's self-written notes at ~/.claude/projects/<project>/memory/
• CLAUDE.md files — User-written instructions at multiple hierarchy levels

Both load into context at session start. More specific instructions take precedence over broader ones.

---

Memory Hierarchy

Memory locations load in priority order (highest priority first):

1. Managed policy — Organization-wide (deployed by IT)
   • macOS: /Library/Application Support/ClaudeCode/CLAUDE.md
   • Linux: /etc/claude-code/CLAUDE.md
2. User memory — Personal, all projects: ~/.claude/CLAUDE.md
3. User rules — Personal modular rules: ~/.claude/rules/*.md
4. Project memory — Team-shared: ./CLAUDE.md or ./.claude/CLAUDE.md
5. Project rules — Modular project rules: ./.claude/rules/*.md
6. Project local — Personal project-specific: ./CLAUDE.local.md (auto-gitignored)
7. Auto memory — Claude's notes: ~/.claude/projects/<project>/memory/MEMORY.md (first 200 lines)
8. Child directory — Nested .claude/CLAUDE.md files load on demand when Claude reads files in those directories

Loading behavior: Files in the directory hierarchy above the working directory load in full at launch. Files in child directories load on demand. Auto memory loads only the first 200 lines of MEMORY.md.

---

Decision Tree — Which Memory Location?

flowchart TD
    Start([Where should this instruction go?]) --> Q1{Who needs it?}
    Q1 -->|Everyone in org| Managed["Managed policy<br>/etc/claude-code/CLAUDE.md"]
    Q1 -->|Just me, all projects| User["User memory<br>~/.claude/CLAUDE.md"]
    Q1 -->|My team on this project| Q2{Single file or modular?}
    Q1 -->|Just me on this project| Local["Project local<br>./CLAUDE.local.md"]
    Q2 -->|Single file| Project["Project memory<br>./CLAUDE.md or ./.claude/CLAUDE.md"]
    Q2 -->|Separate topics| Rules["Project rules<br>./.claude/rules/*.md"]

---

CLAUDE.md Files

Writing Effective Instructions

• Be specific — "Use 2-space indentation" over "Format code properly"
• Use bullet points grouped under descriptive markdown headings
• Include frequently used commands (build, test, lint)
• Document code style, naming conventions, architectural patterns

Bootstrap Project Memory

/init

Generates a starter CLAUDE.md for the current project.

Imports

CLAUDE.md files can import additional files using @path/to/import syntax:

See @README for project overview and @package.json for npm commands.

# Additional Instructions
- git workflow @docs/git-instructions.md

Import rules:

• Relative paths resolve relative to the file containing the import (not cwd)
• Absolute paths and ~ paths supported
• Max recursion depth: 5 hops
• Not evaluated inside code spans or code blocks
• First encounter in a project triggers an approval dialog

Cross-worktree sharing: Use home-directory import so all worktrees share personal instructions:

# Individual Preferences
- @~/.claude/my-project-instructions.md

Edit Memory Files

Use /memory during a session to open any memory file in your system editor. This includes CLAUDE.md files, rules, and auto memory.

---

Modular Rules (.claude/rules/)

For larger projects, organize instructions into separate focused files instead of one large CLAUDE.md.

Structure

.claude/
├── CLAUDE.md              # Main project instructions
└── rules/
    ├── code-style.md      # Code style guidelines
    ├── testing.md         # Testing conventions
    ├── security.md        # Security requirements
    └── frontend/
        ├── react.md       # React-specific rules
        └── styles.md      # CSS/styling rules

All .md files in .claude/rules/ are automatically loaded as project memory. Files are discovered recursively through subdirectories. Symlinks are resolved (circular symlinks handled gracefully).

Path-Specific Rules (Conditional Loading)

Scope rules to specific files using YAML frontmatter with the paths field:

---
paths:
  - "src/api/**/*.ts"
---

# API Development Rules

- All API endpoints must include input validation
- Use the standard error response format

Rules without a paths field load unconditionally.

Supported glob patterns:

• **/*.ts — All TypeScript files in any directory
• src/**/* — All files under src/
• *.md — Markdown files in project root only
• src/components/*.tsx — React components in specific directory
• src/**/*.{ts,tsx} — Brace expansion for multiple extensions
• {src,lib}/**/*.ts — Brace expansion for multiple directories

Multiple patterns:

---
paths:
  - "src/**/*.ts"
  - "lib/**/*.ts"
  - "tests/**/*.test.ts"
---

User-Level Rules

Personal rules at ~/.claude/rules/ apply to all projects. Project rules have higher priority than user rules.

---

Auto Memory

Claude's self-written notes that persist across sessions. Unlike CLAUDE.md (instructions you write for Claude), auto memory contains notes Claude writes for itself.

Storage Location

Each project gets its own memory directory at ~/.claude/projects/<project>/memory/. The <project> path derives from the git repository root. Git worktrees get separate memory directories. Outside git repos, the working directory is used.

~/.claude/projects/<project>/memory/
├── MEMORY.md          # Index file — first 200 lines loaded at session start
├── debugging.md       # Topic file — loaded on demand
├── api-conventions.md # Topic file — loaded on demand
└── ...

How It Works

• First 200 lines of MEMORY.md load into system prompt at session start
• Topic files (e.g., debugging.md) load on demand when Claude needs them
• Claude reads and writes memory files during the session
• Keep MEMORY.md concise — move detailed notes to topic files

Managing Auto Memory

• Edit files directly — they are standard markdown
• Use /memory to open the file selector
• Tell Claude directly: "remember that we use pnpm" or "save to memory that API tests require Redis"
• Ask Claude to forget: "stop remembering X" — Claude will find and remove the entry

Environment Variable Control

export CLAUDE_CODE_DISABLE_AUTO_MEMORY=1  # Force off
export CLAUDE_CODE_DISABLE_AUTO_MEMORY=0  # Force on
# When unset: follows gradual rollout

---

Additional Directory Memory

The --add-dir flag gives Claude access to additional directories. By default, CLAUDE.md files from those directories are NOT loaded. To also load their memory:

CLAUDE_CODE_ADDITIONAL_DIRECTORIES_CLAUDE_MD=1 claude --add-dir ../shared-config

---

Common Tasks

Set Up Memory for a New Project

1. Run /init to bootstrap CLAUDE.md
2. Add build/test/lint commands
3. Document code style, architecture patterns, naming conventions
4. For monorepos, add nested .claude/CLAUDE.md in subdirectories

Organize a Growing CLAUDE.md

When a single CLAUDE.md becomes unwieldy:

1. Create .claude/rules/ directory
2. Split by topic: code-style.md, testing.md, api-design.md
3. Use paths frontmatter for language/directory-specific rules
4. Keep main CLAUDE.md for high-level project overview

Share Rules Across Projects

Symlink shared rules:

# Symlink a shared rules directory
ln -s ~/shared-claude-rules .claude/rules/shared

# Symlink individual rule files
ln -s ~/company-standards/security.md .claude/rules/security.md

Debug Memory Loading Issues

1. Run /memory to see which files are loaded
2. Check file locations match the hierarchy
3. Verify paths frontmatter glob patterns match target files
4. Check import @ references resolve correctly (relative to containing file)
5. For auto memory: verify MEMORY.md is under 200 lines for guaranteed loading

Detailed Reference

For comprehensive details on all memory types, import behavior, glob patterns, and configuration options, see references/memory-reference.md.

SOURCE: Claude Code Memory Documentation (accessed 2026-02-17)