Anthropic has released memory support for Claude Managed Agents, now available in public beta.
The feature lets agents learn from past sessions and share knowledge with other agents. Under the hood, memories are stored as files on a filesystem, using the same bash and code execution tools already built into Managed Agents.
A few things worth noting about how it works:
<ul>
<li>Multiple agents can read from the same memory store at the same time without overwriting each other</li>
<li>Permissions can be scoped at the enterprise level (read-only) or per-user (read-write)</li>
<li>Every memory write is logged as a session event in the Claude console, with rollback and redaction available</li>
<li>Memories can be exported and managed via API</li>
</ul>
Managed Agents are accessible through Anthropic Workbench. There’s also speculation that Conway, Anthropic’s internal agent, will get memory support as well.

Collections

Anthropic has launched memory support for Claude Managed Agents in public beta. Memories are stored as files on a filesystem using existing bash and code execution tools. Key capabilities include concurrent multi-agent reads from a shared memory store, enterprise-level and per-user permission scoping, full audit logging with rollback and redaction, and API-based memory export and management. The feature is accessible via Anthropic Workbench, with speculation that Anthropic's internal agent Conway may also receive memory support.

Claude Managed Agents get memory support in public beta