Show HN: Mwe-MCP – self-hosted memory for AI agents that knows who may know what (github.com)

A new AI framework, mwe-MCP, has been introduced as a self-hosted, agent-agnostic Model Context Protocol server designed to provide all AI agents with structured, persistent memory. This framework allows for the creation of a human-readable Markdown wiki that organizes information in a way that users can easily navigate, edit, and manage. Unlike traditional AI memory systems that rely on opaque vector indices, mwe-MCP offers a user-friendly ecosystem where each fact is individually governed, complete with access control, validity tracking, and a self-organizing mechanism that enhances its functionality over time. The significance of mwe-MCP lies in its ability to offer multi-user, secure memory management tailored for diverse needs—from households to teams—by preserving attribution and access rights dynamically. This enables users to interact with their AI agents without fearing data loss or misinformation, as facts can be marked as expired or retracted without deletion. Furthermore, the system's design emphasizes continuity; it avoids traditional context-reset issues by maintaining a complete structured memory throughout interactions, leading to more accurate recall. With built-in nightly organization and a seamless integration process, mwe-MCP stands to significantly enhance how AI agents manage information, making it more efficient and user-centered in an increasingly collaborative environment.