ME News report, April 24 (UTC+8): According to monitoring by Beating, Anthropic has launched public beta access for built-in memory functionality in Claude Managed Agents—the cloud-based agent hosting platform introduced on April 8. Agents can now learn and accumulate experience across sessions without developers needing to build their own memory infrastructure. Memory is mounted as files within the agent’s filesystem, and Claude reads and writes them directly using its bash and code execution capabilities, treating them identically to other files. Multiple agents can share the same memory storage with granular permission controls: for example, an organization-wide memory repository can be set to read-only, while individual memory repositories allow read-write access, and concurrent operations by multiple agents on the same storage will not overwrite each other. Memory can be exported and managed via API, with all changes logged in an audit trail traceable to specific agents and sessions, supporting rollback and content deletion. The company has shared four customer case studies: Rakuten’s long-running agent learned across sessions to avoid repeating errors, reducing first-time errors by 97%, cutting costs by 27%, and lowering latency by 34%. Wisedocs used cross-session memory to identify recurring document issues in its validation pipeline, improving validation speed by 30%. Netflix’s agent retained insights and human corrections from multi-turn conversations across sessions, eliminating the need for manual prompt and skill updates. Ando leveraged memory to capture collaboration patterns unique to each organization, avoiding the need to build custom memory infrastructure. (Source: BlockBeats)
Claude Managed Agents Introduces Cross-Session Memory; Rakuten Achieves 97% Error Reduction
KuCoinFlashShare
Summary
Anthropic has introduced cross-session memory to Claude Managed Agents, a cloud-based agent hosting platform. Derived from MetaEra, this feature enables agents to retain experience across sessions without requiring custom memory systems. Files are stored in the agent’s file system and accessible via bash or code. Rakuten reported a 97% reduction in first-time errors, 27% cost savings, and 34% lower latency. Other clients, including Wisedocs, Netflix, and Ando, have also shared positive results. New token listings and crypto news continue to underscore advancements in AI-driven infrastructure.
Source:Show original
Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information.
Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.