AutoMem

Research framework for LLM agents that turns memory management into a trainable skill by making file-system memory operations first-class actions and using meta-LLM loops to optimize the memory scaffold and train a dedicated memory specialist.