Anthropic has added persistent memory to Claude Cowork, its agentic work tool, so the app can remember what you told it instead of starting cold every session.
According to SD Times, the same memory Anthropic already used in Claude chat is now built into Cowork, letting users pick up where they left off. Android Authority frames the practical payoff simply: you spend less time re-explaining yourself.
The bigger change is that the two memories are not separate. Android Authority reports that Cowork shares the same memory as chat, meaning context you establish in one place carries into the other. Ukrainian outlet Межа describes this as Anthropic unifying Claude chat and Cowork memory across work tasks.
On the settings that matter most to cautious users, The Next Web reports that the merged memory is on by default, according to the company, while sensitive topics are opt-in. That is a meaningful distinction: the convenience is automatic, but Anthropic is drawing a line around the categories of information it will only retain if you say so.
TechCrunch's framing — that Cowork "finally" remembers what you told the app in chat — captures the mild exasperation behind the update. Cowork was pitched as a tool for doing real work, and a tool that forgets your project, your preferences, and your file structure between sessions creates friction every time you open it.
Why it matters: memory is what turns a chatbot you re-brief every morning into an assistant that actually accumulates context about your work — and by turning it on by default while making sensitive topics opt-in, Anthropic is making a bet about how much users want their AI to remember, and where they want it to forget.