hmem

hmem is a memory infrastructure layer that enables AI agents to retain knowledge across sessions, devices, and AI providers. It solves the problem of agents starting from zero each conversation by organizing memories hierarchically—similar to how humans recall information in chunks and drill down into details on demand. The framework offers token-efficient lazy loading, portability across Claude, Gemini, GPT and local models, and automatic session capture with zero re-read overhead.

Designed for developers building AI agents, hmem works as an MCP (Model Context Protocol) server that agents can install and use autonomously. The maker says it was built with Claude Code. It includes setup guides for AI agents and integrates with tools like OpenClaw, allowing agents to read and write memories without explicit instruction.