Anthropic just fixed one of the most grating parts of using Claude for real work: the wall between its chat interface and Cowork, the agentic side of the product. As of Tuesday, the two share a single memory system. What Claude learns in a casual chat is now available when you switch over to Cowork to actually execute. No more re-explaining your project, your team size, or the conference details you’ve already discussed three times.
The change means Claude behaves like one continuous assistant instead of two separate products wearing the same logo. Anthropic’s example is telling: if you’ve been chatting about an upcoming conference, Cowork will already know the headcount, city, and speaker lineup when you ask it to draft an agenda. Previously, that context would have been lost the moment you switched modes.
There’s also a shift in how memory gets written. Instead of summarizing a conversation at the end, Claude now adds topics to memory as you chat. That makes the handoff between chat and Cowork faster and more fluid, even mid-conversation. You can also see what Claude has retained, edit it, or delete specific memories whenever you want.
Privacy gets a clearer line too. By default, Claude won’t store sensitive categories like health data, race, religion, politics, or gender identity unless you explicitly toggle on “include sensitive topics in memory.” Certain things — Social Security numbers, government IDs, criminal history, immigration status — are never saved, full stop. The feature is live now on Free, Pro, and Max plans across web, desktop, and mobile, though iOS and Android users need to update their apps. Anthropic is essentially betting that persistent, transparent memory is what separates a useful agent from a chatbot you have to constantly re-educate — and honestly, it’s a bet that should have been made sooner.
