Anthropic has announced a significant update to Claude’s memory system, merging the memory used across chat and Claude Cowork so the AI assistant retains context regardless of which product a user engages with. The change, revealed Tuesday, addresses a common frustration among agent users: repeatedly rebriefing the AI on information it should already know.
The company said Claude will now surface what it has stored in memory, letting users read, edit, or delete information on any topic. According to Anthropic, the update lets users move more fluidly between research and action, since Cowork will already understand context built up during earlier chats, such as project details, headcounts, locations, or speaker lists, without requiring users to re-enter that information.
Rather than summarizing conversations only once they end, Claude will now add topics to memory continuously as conversations happen, allowing updates to sync faster between chat and Cowork even mid-conversation.
Anthropic stated that Claude will not store sensitive personal information, including health data, race, religion, and political or gender identity, by default, though users can opt in via a toggle. Certain data, including government IDs and criminal history, will never be saved. The feature is now enabled by default across Free, Pro, and Max plans.
