Anthropic has updated Claude’s memory system to allow information shared in regular chats to carry over to Claude Cowork and vice versa. The move is aimed at reducing the need for users to repeatedly brief the AI when moving between conversations and tasks.
With the update, Claude uses a shared memory across Chat and Cowork. This means details discussed during an earlier conversation, such as project priorities, deadlines or preferences, can be available when users later hand a task to Cowork. Information picked up during a Cowork task can also feed back into future chats.
Anthropic has also changed how Claude updates its memory. Rather than waiting until a conversation ends to create a summary, Claude can now add relevant topics to memory while users are still chatting. Users can pause or reset memory when they choose.
Another major change is greater visibility into what Claude has retained. Saved memories are organised by topic, allowing users to view, edit or delete individual entries. This gives users more direct control over the information that carries across their interactions with Claude.
Anthropic says sensitive subjects such as health, political views, religious beliefs, race and ethnicity are not stored by default. Users can choose to enable sensitive topics in Memory settings and Claude will notify them when such a topic is saved. Certain information, including government-issued identification, Social Security numbers, criminal history and immigration status, will not be stored in memory.
The memory update is available by default on Free, Pro and Max plans across Claude’s web, desktop and mobile experiences, with mobile users required to update to the latest version of the app.
The change comes as Anthropic continues to develop Claude as an agentic work assistant rather than simply a chatbot. Cowork can handle multi-step tasks and work with files, browsers and connected applications, making persistent context increasingly important as users move from asking Claude questions to handing it ongoing work. Anthropic’s own safety guidance notes that Cowork sessions can run in the cloud and interact with files, browsers and apps, making control over access and retained information particularly relevant.
The update effectively brings Chat and Cowork closer to functioning as one continuous Claude experience, allowing users to move from brainstorming and research to execution without starting from scratch each time.