Show HN: Spomory – graph-based MCP memory with an RL-trained forgetting policy (spomory.yliuai.com)

A new tool called Spomory introduces a graph-based memory system that allows for seamless context sharing across multiple AI platforms. Designed to eliminate the repetitive onboarding process when switching between different AI tools, Spomory enables users to establish a shared memory that any AI can access. This means if you provide information such as your role or project details in one AI, other AIs will recognize and retain that context, potentially enhancing productivity and continuity across sessions. The significance of Spomory lies in its architecture and how it retains memory. Unlike conventional AI memory systems that operate within isolated applications, Spomory uses a dual-layer memory graph that organizes facts into structured triples. This allows for efficient retrieval and multi-hop connections, ensuring that related information can be accessed contextually. A reinforcement learning-trained forgetting policy determines how memories are added, updated, or retired, contributing to more accurate memory management and user control over data privacy. Moreover, Spomory operates in both local and cloud-hosted environments, making it flexible and user-friendly, with robust data handling features like exportability and deletion options.