Locus

Locus is a hierarchical, markdown-based memory system specifically engineered for autonomous AI agents, drawing inspiration from the 'Method of Loci'. Its fundamental principle is to minimize AI context window usage by dynamically loading only the 'rooms' (directories with specific knowledge) an agent needs, rather than the entire 'palace' of information. Agents navigate a central `INDEX.md` to locate relevant knowledge, read specific files, and record session logs that can later be consolidated into canonical files. This system significantly reduces context line usage compared to flat memory approaches while ensuring comprehensive recall and providing a structured framework for knowledge management. It integrates via the Model Context Protocol (MCP) and offers robust security features.

主要功能

使用案例