A pragmatic, low-overhead Model Context Protocol (MCP) server that provides Large Language Models (LLMs) with a persistent, local long-term memory.