Anthropic Unifies Memory Across Claude and Cowork Services
The AI company has linked user data between its chatbot and collaborative workspace, though global privacy controls remain available.
Anthropic has implemented a shared memory system between its Claude AI and Cowork services to create a more seamless user experience. This integration allows information learned in one service to be accessible in the other, effectively unifying the AI's knowledge of a user across different platforms.
According to reports from The Register and ZDNET, the two services now draw from a single memory pool. This means that preferences, facts, or context established during a general chat with Claude can be utilized within the Cowork collaborative workspace, and vice versa. However, this synchronization does not extend to the entire ecosystem; Claude Code remains isolated from this shared memory system for the time being.
The Push for Persistence
This move is part of a broader effort by Anthropic to expand the persistence of user data across its ecosystem. By bridging the gap between its general-purpose AI and its specialized workspace tool, the company aims to reduce the need for users to repeat instructions or provide the same background information when switching between different Claude-powered environments.
Privacy and Control
The unification of these memory pools has raised questions regarding granular data control. While the system is unified, users do not have the option to split the memory pool specifically between Claude and Cowork while keeping the feature active for both.
To address privacy concerns, Anthropic provides global controls. Users can pause, reset, or toggle off memory generation from chats entirely. Additionally, users have the ability to edit or delete specific memory topics to refine what the AI remembers. Despite these tools, the lack of a "split" option means professional data in Cowork and personal interactions in Claude are stored in the same repository unless the feature is disabled globally.
What's Next
As AI agents move toward more autonomous operation, the ability to maintain a consistent identity and knowledge base across tools is a key technical hurdle. Industry observers will be watching to see if Anthropic introduces more nuanced permissions—such as "work-only" or "personal-only" memory silos—to satisfy enterprise security requirements. For now, the boundary remains between the chat/workspace ecosystem and the isolated environment of Claude Code.