Anthropic has updated Claude’s memory system, addressing a common frustration where users had to repeatedly inform the AI about previously discussed topics. The company revealed on Tuesday that it is integrating the memory across Claude’s chat and Claude Cowork functionalities, ensuring that information learned in one interaction is retained and accessible in others.
Additionally, Anthropic is providing users with visibility into Claude’s stored memories, giving them the ability to review, modify, or remove information on any subject.
This enhancement is set to streamline the user experience, enabling smoother transitions for Claude’s users between the stages of research and practical application.
Previously, users might have conceptualized project ideas through ongoing chats with Claude, only to face the tedious task of re-entering all relevant details into Cowork when it was time to implement those ideas. This cumbersome process has been resolved. With this update, Claude now operates as a single, cohesive assistant, eliminating the prior distinction between its separate chat and Cowork components.
Anthropic illustrates this with scenarios such as drafting a manager update or a conference agenda. Cowork will now recall pertinent information from previous discussions, such as conference attendees, location, and speakers, thereby eliminating the need for users to re-enter these details.
Furthermore, Claude will now integrate topics into its memory incrementally during a chat, rather than just summarizing the entire conversation upon completion. This facilitates easier transitions between chat and Cowork, as memories are updated and shared more rapidly across both experiences, even mid-conversation.
Anthropic states that, by default, Claude will refrain from storing personal or sensitive data such as health information, racial or ethnic background, religious or political affiliations, and gender identity. Users retain the option to enable the storage of such information by activating the “include sensitive topics in memory” setting. The application will also issue notifications when any sensitive topic is recorded in memory.
Nonetheless, the company confirms that Claude will never store certain categories of information, including government-issued identification, Social Security numbers, criminal records, immigration status, or any other data that contravenes its acceptable use policy.
This memory functionality is automatically active for Free, Pro, and Max subscribers across web, desktop, and mobile platforms. iOS and Android users will be required to update their applications to the most recent version to access it.