On the heels of its recent relesae of Claude Managed Agents, Anthropic this week has released to public beta Memory on Claude Managed Agents, which enables agents to learn from past sessions and share what they’ve learned with other agents. The memories are stored as files so developers can keep control over what these agents retain.
In a blog announcing the memory capability, Anthropic wrote: “We’ve found that agents are most effective with memory when it builds on the tools they already use. Memory on Managed Agents mounts directly onto a filesystem, so Claude can rely on the same bash and code execution capabilities that make it effective at agentic tasks. With filesystem-based memory, our latest models save more comprehensive, well-organized memories and are more discerning about what to remember for a given task.”
According to the company, Memory is. built on the same bash and code execution tools that are already used in Claude Managed agents; those agents can access and work against the same store concurrently without overwriting each other. Also, it noted that scoped permissions enable control over agents, saying that stores used across an enterprise might be set as read-only, while per-store users might allow read write.
Memories can also be exported and managed via API as a means for developers to retain full control. In the announcement, it wrote that every write becomes as session event in the Claude console, so teams can trace what the agent learned and where it came from, with the ability to roll back or redact if something went awry.
Visit the Claude Console or use the new CLI to deploy your first agent with memory. Explore the documentation to learn more.
