Mnemory – Persistent memory for AI agents (github.com)

Mnemory has announced a new persistent memory solution for AI agents, allowing them to remember user preferences, facts, and contexts beyond individual conversations. This self-hosted MCP server integrates seamlessly with various AI models like Claude Code and ChatGPT, enabling immediate personalization without the need for changes to system prompts. Mnemory emphasizes data security, as all memories are stored locally, eliminating cloud dependencies and unauthorized access. The significance of mnemory lies in its advanced processing capabilities, utilizing a unified large language model (LLM) for efficient fact extraction, classification, and contradiction resolution in a single step. Its dual-tier memory system ensures both quick access to summarized content and detailed artifact storage, while features like semantic search and built-in health checks maintain memory integrity over time. Designed for easy deployment with minimal configuration, mnemory can be employed by AI developers to enhance user experiences significantly, fostering more personalized interactions and permanent memory features in various MCP-compatible clients.