claude-memory-3layer

To add persistent memory to Claude Code sessions, use a three-layer architecture combining a knowledge graph, daily notes, and tacit knowledge for structured retention and retrieval across interactions.

Yes, you can track Claude Code memory changes with git because all memory entries are stored directly within the project directory, making them fully accessible for standard version control workflows.

Claude Code memory systems manage token consumption by enforcing a configurable token budget per topic, ensuring precise control over context window usage during knowledge retrieval and storage.

Yes, you need Python 3.8 or higher to execute the scripts required by a hook-driven memory system, and Claude Code v2.1.0 or higher is necessary to support the lifecycle hooks.

Outdated knowledge entries get cleaned up through automated smart review processes that assign Active or Superseded statuses, tracking the lifecycle of knowledge within the memory system.

No, a custom three-layer memory system is designed to coexist with official auto-memory features, complementing them without conflict to enhance overall knowledge management capabilities.