Home→Glossary→CLAUDE.md

CLAUDE.md

CLAUDE.md

The project file Claude Code reads automatically before any work starts

✓ Building Blocks
Talk to AY Automate ↗
Read automaticallyProject-scopedFolds into the system promptNot a per-message instruction

What is a CLAUDE.md

CLAUDE.md is a markdown file placed in a repository that Claude Code reads automatically the moment a session starts there, before the first user message ever arrives. It typically holds the things a human collaborator would otherwise have to explain every single time: naming conventions, which directories are off limits, how commits should be structured, which commands actually run the test suite.

Its value is in not having to repeat any of that. Without a CLAUDE.md, a project's conventions live only in a person's head, or get retyped into the first message of every new session. With one, that context loads automatically and applies to every session in the repository, whether it is a five-minute fix or an hours-long build.

CLAUDE.md content is folded into the system prompt layer rather than sitting in the visible conversation, which is why it feels standing rather than like an instruction the user typed. It is scoped to the repository it lives in, a monorepo can carry more than one, each applying to its own part of the project, rather than one global file governing everything.

AY Automate

Build your own agent skills, not just prompts.

We design, build, and ship the Claude Code skill libraries production teams actually run on.

Talk to AY Automate ↗

Key properties

✓Loaded once at session start, not re-sent with every message
✓Scoped to the repository it lives in, not global by default
✓Folded into the system prompt rather than treated as a chat message