Anthropic merges Claude Chat and Cowork memory

Anthropic is changing how Claude remembers information across its products. TechCrunch reported on August 25, 2026 that the company is merging the memory system used by Claude’s chat experience with the one used by Claude Cowork. Information learned in a conversation can therefore be available when a user switches to Cowork to take action, instead of requiring the user to repeat the context. The change addresses a practical limitation of agent software. A person might discuss a project, collect ideas or plan an event in Claude Chat, then open Cowork to draft a document, prepare an agenda or complete another task. Previously, Cowork could behave as if that earlier discussion had never happened. Anthropic’s example is a conference: Claude could retain details such as the headcount, city and speakers while helping prepare follow-up work. Anthropic is also changing when memories are created. Rather than waiting until a conversation ends and then summarising it, Claude can add topics to memory while the chat is still in progress. The company says users can inspect what Claude has retained, edit it or delete it. That makes memory more useful for continuity, but it also makes the contents of the memory store an important part of the user experience. The feature is enabled by default for Free, Pro and Max plans on the web, desktop and mobile. iOS and Android users need the latest app version. Anthropic says Claude will not store personal or sensitive topics such as health data, race, religion, politics or gender identity by default. Users can opt in to sensitive topics, and the app will notify them when one is saved. The company says it will never save items such as government-issued identification, Social Security numbers, criminal history or immigration status. For users, the benefit is less context-switching and fewer repeated explanations. For businesses and makers, the update highlights a harder design question: persistent memory must be useful without becoming an invisible record of everything a person has said. The ability to review and delete memories is therefore as important as the convenience. Anthropic’s announcement, as reported by TechCrunch, describes the product behaviour and its stated boundaries; teams should still check settings, retention policies and account controls before using memory with confidential work. The wider market signal is that AI assistants are becoming continuous collaborators rather than isolated chat windows.