Mengram

Mengram is an open-source AI memory engine designed to give applications human-like long-term memory and personalization. It helps developers structure, store, and retrieve user-related information so AI agents can remember facts, past interactions, and workflows over time. The primary purpose of Mengram is to provide a robust, extensible memory layer that can be integrated into AI systems, chatbots, and automation tools to deliver more consistent and context-aware behavior.

At its core, Mengram models three human-inspired memory types: semantic memory for storing factual knowledge, episodic memory for capturing events and interactions, and procedural memory for representing workflows, processes, and step-by-step tasks. A built-in Cognitive Profile enables instant personalization by aggregating user traits, preferences, and behavioral patterns, allowing AI agents to adapt responses and decisions to each individual. The system supports structured querying and retrieval, enabling applications to reference prior sessions, reuse learned procedures, and maintain continuity across conversations or tasks. As a free and open-source project, Mengram can be self-hosted, extended, and audited, making it suitable for teams with strict security or compliance requirements.