Alibaba Cloud upgrades Memory service for Personal Agent long-term memory

On November 2, 2026, IT之家 reported that Alibaba Cloud had announced a recent upgrade to its Memory service. The upgrade targets Personal Agent scenarios, strengthening the service's ability to remember user preferences, plans, and past commitments. Access is offered through MCP and CLI, among other channels. According to IT之家, Memory extracts key information from conversations and, following defined rules, stores it as memories and user profiles. In subsequent sessions, an Agent retrieves relevant memories based on user identity and the current question, adds them to the task context, and uses that context to generate responses. New valid information can continue to be written into memory so the Agent keeps tracking user needs and asks fewer repetitive questions. In a recent HaluMem-medium comparison, Memory achieved the highest scores on all four reported measures: recall, precision, and F1 for memory extraction, plus the memory QA C metric. In the PersonaMem-v2 evaluation, it ranked first among the eight listed systems with an overall score of 56.82, and posted the top score on seven of nine task dimensions. As an example in financial advisory, when a user says they prefer conservative investments, may need the money within six months, and want to avoid large fluctuations, Memory can save risk preference, funding horizon, and liquidity needs as a user profile. In later sessions, an Agent can call on that profile and combine it with product information and business rules to compare options and issue risk reminders. The publicly available record stops here; the report does not cover the service's launch timing, regions, or pricing.