TMTPOST — Tencent Cloud on Thursday released Agent Memory 2.0.0, extending long-term memory capabilities from individual agents to full team collaboration. The upgrade centers on Team Memory, which converts code knowledge, project documents, historical conversations and working methods generated during agent-team interactions into shared, reusable assets. These assets can be governed by ownership, version and visibility rules, then assembled on demand according to each agent’s role and specific task. The open-source system, available on GitHub under the TencentCloud organization, organizes memory into four asset types—Chat Memory, Skill, Wiki and CodeGraph—and has already climbed to the top of GitHub’s TypeScript trending list with more than 15,000 stars. The release builds on earlier individual-memory features that reduced token consumption by compressing context and preserving structured task states. By enabling agents to inherit team experience across sessions and frameworks, the platform aims to cut repeated explanations and improve consistency in multi-agent enterprise workflows spanning software development, documentation and operational processes.