Alibaba cloud enhances memory service for personal agents, boosting contextual recall and personalization

Alibaba Cloud has upgraded its Memory service, specifically targeting Personal Agent applications to improve the ability to retain user preferences, plans, and historical agreements. This enhancement allows the Agent to retrieve relevant memories during subsequent conversations, integrating them into the task context to generate more informed responses and reduce redundant questioning. Performance benchmarks indicate significant improvements; in recent comparisons, Memory achieved the highest scores across recall rate, accuracy, F1, and memory Q&A C metrics when compared against HaluMem-medium. Furthermore, in the PersonaMem-v2 evaluation, Memory ranked first among eight systems with an overall score of 56.82, securing the top spot in seven out of nine task dimensions. Developers gain granular control, allowing them to define custom memory rules and profile fields via the console for management and tracking. Integration is supported through APIs for writing and semantic retrieval, as well as via MCP and CLI to connect with frameworks like Codex, DeepSeek Harness, Claude Code, and OpenClaw. The service can also collaborate with Skills, enabling reusable process flows that dynamically adapt based on the user's latest profile data.