Show HN: Cortex – local-first encrypted memory for AI agents (Rust, MCP) (github.com)

Cortex has launched a local-first, encrypted memory solution designed specifically for AI agents, enabling on-device data storage without relying on cloud services. This innovative tool, implemented in Rust and running entirely client-side, allows users to manage their AI's memory securely. With a minimal footprint of 124KB as a WASM application and a 3.8MB binary, Cortex eliminates delays and costs associated with cloud memory solutions by offering sub-millisecond latency for memory operations and completely removing third-party data handling. This advance is significant for the AI/ML community as it challenges existing memory solutions, which often expose users' data to external servers and long response times. Cortex’s architecture supports a four-tier memory system that incorporates working, episodic, semantic, and procedural memory, along with features like Bayesian self-correcting understanding and cross-platform identity resolution. The ability to persistently associate memories across different contexts while maintaining privacy marks a critical step forward in developing user-centric AI, catering not just to performance but also to privacy needs in an increasingly data-sensitive world.