Anthropic has merged the memory systems behind Claude chat and its Cowork agent app, ending the need to re-explain project details every time a task moves between the two.
The change means context built up over months of conversation carries into Cowork the moment a task is handed over, and anything learned in Cowork flows back into chat. Anthropic is also making memory inspectable, letting users read, edit, or delete what Claude has stored on any topic.
Previously, ideas developed in chat had to be restated when a user switched to Cowork to act on them. Now Claude adds topics to memory continuously during a conversation rather than summarizing when it ends, so the two experiences stay in sync even mid-chat.
By default, the assistant will not store sensitive categories such as health data, race, ethnicity, religion, politics, or gender identity, though users can enable them with an “include sensitive topics in memory” toggle. Claude will also notify users whenever it saves a sensitive topic. Government-issued IDs, Social Security numbers, criminal history, and immigration status are never stored.
Memory is on by default across Free, Pro, and Max plans on web, desktop, and mobile, with iOS and Android users needing the latest app update. Claude Code is not part of the merged memory system.
