We Read Claude Code's Leaked Source. Here's How Its Memory Actually Works

This post analyzes the memory architecture of Claude Code, identifying limitations such as a 200-line index cap, grep-only retrieval, lack of reasoning capture, complexity stacking, and memory lock-in. It then introduces 'memsearch', a persistent memory layer for AI coding agents that utilizes Milvus for hybrid semantic and full-text search. Memsearch stores memory as dated Markdown files, with Milvus providing a rebuildable vector index for efficient recall. It employs semantic search, BM25, and RRF for robust memory retrieval, addressing the limitations of Claude Code's native memory system.