Anthropic has launched a memory feature for Claude that enables the AI assistant to remember users' and teams' projects, preferences, and work patterns. The feature is rolling out to Team and Enterprise plan users starting September 11, 2025, with Enterprise admins able to disable memory for their organiSation at any time.
Claude's memory system focuses on professional context, remembering team processes, client needs, project details, and priorities. Sales teams can maintain client context across deals, product teams preserve specifications across sprints, and executives track initiatives without rebuilding context. The feature creates separate memory instances for each project when users utilise Claude's projects functionality, ensuring product launch planning remains separate from client work and confidential discussions stay separate from general operations.
Users can view and edit Claude's memory through a memory summary accessible in settings. Based on user instructions about what to focus on or ignore, Claude adjusts which memories it references. The system allows users to update the memory summary at any time through conversation with Claude.
Anthropic has introduced Incognito chats alongside the memory feature, available to all Claude users. These conversations don't appear in conversation history or save to memory, providing a clean slate for sensitive brainstorming, confidential strategy discussions, or situations requiring fresh conversation without previous context. For Team and Enterprise users utilising Incognito mode, standard data retention settings apply.
The company states memory introduces new safety considerations and has designed the feature for work settings while avoiding sensitive conversations and topics. Anthropic is taking a phased approach to ensure responsible deployment of these capabilities and will continue evaluating how memory functions across different Claude usage patterns before expanding availability.
Enterprise administrators maintain control over memory activation for their organisations. The project-based memory separation serves as a safety guardrail keeping sensitive conversations contained while enabling teams to manage complex, concurrent initiatives without mixing unrelated details across different business functions.
Memory positions Claude as a continuous work companion rather than a session-based tool, potentially increasing user engagement and productivity by eliminating repetitive context setting. The phased rollout approach demonstrates Anthropic's focus on enterprise safety requirements. Success factors include granular user controls and project-based memory segmentation addressing enterprise data governance needs.