Anthropic Launches Memory for Claude Managed Agents in Public Beta
Summary
- • Anthropic releases Memory feature for Claude Managed Agents in public beta
- • Agents can now retain and reuse information across sessions without manual prompt updates
- • Memory is filesystem-based with API management, permissions scoping, and full audit trails
- • Early adopters include Netflix, Rakuten, Wisedocs, and Ando
Details
Anthropic releases Memory for Claude Managed Agents in public beta, available immediately to all users
Memory enables agents to accumulate knowledge across sessions, eliminating the need for developers to manually update prompts with context from prior interactions. Available via Claude Console and programmatic APIs.
Memory is filesystem-based: data stored as exportable files, managed through APIs, with permission scoping
The architecture allows organizations to manage memory through APIs and apply permission scoping to control access at various organizational levels. This prioritizes portability and enterprise control over a black-box approach.
All memory changes are logged with per-session, per-agent audit trails supporting rollback and redaction
Enterprise customers can granularly govern what agents remember and correct or erase memory entries. This addresses compliance and governance concerns for regulated industries deploying autonomous agents.
Early adopters Netflix, Rakuten, Wisedocs, and Ando are using Memory to streamline workflows and reduce errors
The named early adopters span media, e-commerce, healthcare documentation, and fintech — signaling broad applicability across enterprise verticals rather than a niche use case.
Memory release continues Anthropic's push to position Claude as the enterprise-grade agent platform of choice
The combination of persistent memory, auditability, and permission controls directly targets enterprise procurement requirements around transparency and data governance — differentiating Claude from less auditable competitors.
Product Launch = newly released product or feature; New Tech = new technical capability; Industry Update = business or ecosystem development; Strategy = competitive positioning intent
What This Means
Persistent memory is a foundational capability for enterprise AI agents — without it, every session starts from scratch, forcing developers to re-inject context and limiting how much agents can improve over time. By building Memory with filesystem storage, API controls, and full audit trails, Anthropic is directly addressing the governance and compliance requirements that most enterprises cite as blockers to agentic AI adoption. This puts Claude Managed Agents in stronger competition with other enterprise agent platforms and raises the bar for what production-ready agent infrastructure looks like.