Claude Managed Agents Memory Enters Public Beta
Anthropic launched the public beta of Memory on Claude Managed Agents on April 23, 2026. Agents now learn from every session via an "intelligence-optimised memory layer" — memories are stored as files, making them exportable and manageable via the API. Developers retain full control over what is stored, surfaced, and deleted. Beta access is live at claude.com with documentation linked from the announcement.
Why It Matters
Persistent, developer-controlled memory transforms Claude Managed Agents from stateless session tools into continuous-learning agent systems. The file-based approach with full API access means organisations can audit, curate, and integrate agent memory into their own data workflows — a critical requirement for enterprise agentic deployments where context continuity and compliance both matter.