Breaking the long-term memory bottleneck of AI, Alibaba Cloud empowers Personal Agent to move towards deep personalization. On November 2, Alibaba Cloud announced a comprehensive upgrade of its Memory service, focusing on enhancing long-term memory and personalized processing capabilities of user preferences, plans and historical appointments for Personal Agent scenarios, and comprehensively providing

Breaking the long-term memory bottleneck of AI, Alibaba Cloud empowers Personal Agents to move toward deep personalization. On November 2, Alibaba Cloud announced a comprehensive upgrade of its Memory service, focusing on enhancing long-term memory and personalized processing capabilities of user preferences, plans, and historical appointments for Personal Agent scenarios, and fully providing mainstream access methods such as [MCP] and [CLI]. This service extracts key information from conversations in real time and automatically saves it as user portraits and structured memories according to rules. In subsequent interactions, the system can accurately retrieve and call relevant context based on the user's identity and current questions, enabling continuous follow-up of user needs and effectively reducing repeated inquiries. Evaluation data shows that in the recently announced [HaluMem-medium] comparison, Memory achieved the highest scores in the recall and accuracy [F1] indicators of memory retrieval and the memory question and answer C indicator; in the PersonaMem-v2] evaluation, it ranked first among the [8] participating systems with an overall score of 56.82, and ranked first in [7] of the 9 task dimensions. In actual application scenarios, taking financial consulting as an example, the system can automatically summarize the user's risk preference and liquidity needs and generate a portrait, collaborate with [Agent] in subsequent conversations to complete product comparisons and risk reminders, and dynamically update the memory when user preferences change. Currently, developers can not only customize memory rules, portrait fields and track retrieval effects in the console, but can also seamlessly integrate this capability into mainstream [Agent] tools and frameworks such as Codex, DeepSeek Harness, Claude Code] and OpenClaw through API, MCP or CLI. In addition, Memory] can be deeply cooperated with the Skill] module to achieve decoupling and efficient collaboration of individual preferences and reusable task processes. This upgrade marks that personal assistants are accelerating to break through the context bottleneck of traditional single-turn conversations, evolving towards both deep personalization and long-term companionship capabilities, providing solid infrastructure support for [Agents] to penetrate into complex vertical scenarios. via AI News (author: AI Base)