The missing layer in AI tooling: sharing what your assistant already knows

Developer Uri Shmueli has introduced memshare, an open-source tool designed to address the fragmentation of AI memory. Currently, AI assistants store context in isolated silos, forcing users and teammates to start from scratch when switching tools or onboarding new members. Memshare treats AI memory as a portable data type, storing information in local, human-readable JSON files. By utilizing the Model Context Protocol (MCP), it allows developers to capture, manage, and share project-specific knowledge across different AI tools like Claude, Cursor, and Windsurf. The project emphasizes a strict consent model, featuring automated PII scanning and individual item-level import controls to ensure privacy. Because it relies on local files rather than proprietary databases, memshare remains git-friendly and tool-agnostic, aiming to bridge the gap between individual AI workflows and collaborative team environments.