Anthropic connects Claude chat memory with Claude Cowork
Anthropic has merged the memory used in Claude chat with the memory available in Claude Cowork, allowing information learned in one experience to be carried into the other. The change is enabled by default for Free, Pro and Max subscribers on the web, desktop and mobile apps; iOS and Android users must update to the latest app version.
The update is designed to remove a frequent interruption in agent-assisted work: having to restate project context when moving from a conversation to an action-oriented workflow. Claude can now retain details developed during chat and use them when a user opens Cowork.
One memory across chat and Cowork
Anthropic describes the result as a more continuous assistant rather than two separate products. A user who has discussed a conference with Claude, for example, could ask Cowork to prepare an agenda or a manager update without re-entering the headcount, city and speaker information.
The product direction builds on Claude Cowork office-work agent as Anthropic extends the office-work agent with context retained from ordinary conversations. The shared memory is intended to make the transition from research and planning to execution less repetitive.
Memories are updated during conversations
Claude will now add topics to memory as a chat proceeds, instead of only summarising a conversation after it ends. Anthropic says this allows newly stored context to be updated and shared between chat and Cowork more quickly, including while the conversation remains active.
Users are also being given controls to view what Claude has retained, edit it or delete it. Those controls matter because persistent context can be useful for repeated work, but it must remain visible and manageable to the person using the service.
Limits for sensitive information
By default, Claude will not store personal or sensitive information such as health data, race, ethnicity, religious beliefs, politics or gender identity. Users may choose to enable an “include sensitive topics in memory” setting, and the app will notify them when a sensitive topic is saved.
Anthropic says Claude will never retain government-issued identification, Social Security numbers, criminal history, immigration status or information that violates its acceptable use policy. For businesses, the practical implication is to inspect and maintain Claude’s shared memory before relying on Cowork outputs that use long-running project context.