Skip to main content
Claude Code reads plain Markdown files called memory files to load persistent instructions at the start of every session. You write instructions once and Claude follows them automatically — no need to re-explain coding standards, architecture decisions, or workflow preferences every time.

​
Memory file types

Memory files are loaded in the following order (lowest to highest priority):
TypeLocationPurpose
Managed/etc/claude-code/CLAUDE.mdOrganization-wide rules set by administrators
User~/.claude/CLAUDE.mdYour personal global instructions for all projects
ProjectCLAUDE.md, .claude/CLAUDE.md, .claude/rules/*.mdTeam instructions checked into the repo
LocalCLAUDE.local.mdYour private project instructions, not committed
Files loaded later (higher in the list above) have higher priority — Claude pays more attention to them.
Add CLAUDE.local.md to your .gitignore to keep personal preferences out of the shared repository.

​
Project memory

Place a CLAUDE.md at your project root (or .claude/CLAUDE.md) to document conventions Claude should always follow:
CLAUDE.md
# Project conventions

## Code style
- Use TypeScript strict mode
- Prefer `const` over `let`
- No `any` types — use `unknown` and narrow

## Testing
- Tests live in `src/__tests__/`
- Use Vitest, not Jest
- Every new function needs a unit test

## Git
- Commit messages: `type(scope): description`
- Never commit directly to `main`

​
Rules directory

For larger projects, split instructions across multiple files in .claude/rules/: 
.claude/
  rules/
    coding-standards.md
    api-conventions.md
    testing-guide.md
All .md files in that directory are loaded automatically.

​
User memory

Your global user memory at ~/.claude/CLAUDE.md applies to every project. Use it for personal preferences that shouldn’t live in the repo:
~/.claude/CLAUDE.md
- Always explain what you're about to do before making changes
- Prefer verbose variable names
- Ask before refactoring files I haven't mentioned

​
The @include directive

Memory files can include other files using @ notation:
CLAUDE.md
# Main instructions

@./docs/architecture.md
@./docs/api-conventions.md

## Additional notes
...
Supported path formats:
  • @path — relative to the including file
  • @./relative/path — explicit relative
  • @~/home/path — from your home directory
  • @/absolute/path — absolute path
Circular references are automatically prevented. Non-existent files are silently ignored.

​
Managing memory with /memory

Use the /memory command inside a session to open your memory files in your editor: 
> /memory
This opens the relevant CLAUDE.md file so you can add or update instructions without leaving the terminal.

​
What to put in memory files

Language versions, formatting rules, naming conventions, patterns to avoid, preferred libraries — anything Claude should consistently apply across files.
Directory structure, module boundaries, how the build system works, where generated files go, key abstractions and their purpose.
How to run tests, how to build, branch naming, PR process, how to deploy to staging — operational context Claude needs to work efficiently.
Explicit don’ts are often more useful than do’s. List anti-patterns, deprecated APIs, files that should never be modified directly, and commands that require human review.

Build docs developers (and LLMs) love

Get started for freeTalk to us