Welcome — what claude-code-memory is and how to use it

[manja316]

Welcome — what claude-code-memory is and how to use it

Hi 👋 I'm the maintainer. Opening this discussion as the entry point for anyone landing here from PyPI or GitHub search.

What this is: a zero-dependency memory layer for Claude Code (Anthropic's CLI). It gives your CLAUDE.md a structured, persistent memory directory that survives /clear, session restarts, and context compaction — so you stop re-teaching Claude the same facts about your project every morning.

Why it exists: I kept losing the same context — API quirks, debugging conventions, "we tried X, it broke because Y" — every time a Claude Code session ended. The fix is a tiny memory protocol baked into your global CLAUDE.md: typed memories (user / feedback / project / reference), an index file (MEMORY.md) that's always loaded, and discipline about when to save vs. let it live in code.

Install:

pip install claude-code-memory
claude-code-memory init

That drops a memory directory into ~/.claude/ and a CLAUDE.md snippet you paste into your global instructions. From there, Claude writes its own memory files as it learns about you and your projects.

What I'd love to hear in this thread:

1. If you've installed it — what's the first memory Claude saved that surprised you?
2. If you haven't — what kind of context do you find yourself re-explaining every session? (I want to make sure the memory types cover the common cases.)
3. Ideas / gripes / "this would be 10× better if…" — drop them, I read all of them.

PyPI: https://pypi.org/project/claude-code-memory/
Repo: https://github.com/LuciferForge/claude-code-memory

— Manjunath