MemQ introduces a novel approach to enhance memory retrieval in LLM agents by integrating Q-learning with a provenance DAG, allowing for more effective accumulation and use of episodic memories. The method shows significant improvements across various benchmarks, particularly in multi-step tasks, indicating potential advancements in AI memory systems and their applications.