Anthropic merges Claude chat and Cowork memories
The update unifies context across different modes of the AI tool, eliminating the need for users to rebrief the system when moving from research to execution.
Anthropic has announced a merger of the memory systems used by its Claude chat interface and Claude Cowork. According to ZDNet AI, this change creates a two-way flow where information shared in one tool is automatically accessible in the other. The update aims to create a continuous assistant experience, allowing users to move between different tasks without restarting the context.
TechCrunch AI reports that this merger solves a primary friction point for users of agentic tools: the constant requirement to provide repeated briefings. Previously, data developed during research phases in Claude chat remained isolated from Cowork. With the new unified system, if a user brainstorms conference details—such as headcount, city, and speakers—during a chat session, Cowork will already possess that information when it begins drafting logistics documents or budgets.
ZDNet AI notes that while Claude chat and Claude Cowork are now linked, this update does not appear to include Claude Code. When asked about the exclusion of Claude Code, Anthropic representatives stated the current focus remains specifically on Cowork and chat.
Transition from summary to dynamic updates
The technical implementation of memory within these tools is shifting from a post-session process to a real-time one. Previously, Claude would summarize conversations once they were finished to build its memory. According to both ZDNet AI and TechCrunch AI, the system now adds topics to memory dynamically as the user chats.
This shift allows for more immediate synchronization between the two experiences. Because memories are updated during active sessions rather than at the end of a conversation, users can move back and forth between research in chat and task execution in Cowork with more up-to-date context.
Privacy controls and data exclusions
Anthropic has implemented specific defaults to manage how sensitive information is retained. According to TechCrunch AI, Claude will not store personal or sensitive topics by default, including health data, race, ethnicity, religious beliefs, politics, and gender identity. However, users can opt-in to save these types of details by toggling a setting in the memory menu.
If a user chooses to enable sensitive topic retention, ZDNet AI reports that Claude will display a notice whenever such information is being recorded. Despite these user options, Anthropic maintains hard exclusions for certain data types. The company states it will never store government-issued identification numbers, criminal history, immigration status, or any information that violates its acceptable use policy.
Users can manage their retained information through the Memory section of the settings menu. This interface allows users to read, edit, or delete saved memories globally. For example, if a user corrects an old company name in one session, the update applies across all subsequent interactions.
Availability by subscription tier
The availability of these memory features varies depending on the user’s plan. ZDNet AI reports that memory is enabled by default for users on Free, Pro, and Max plans across web, desktop, and mobile platforms. For those on Enterprise or Teams tiers, both the memory features and the sensitive topic options are turned off by default and are subject to eligibility requirements.
Mobile users on iOS and Android will need to update their applications to the latest version to access these changes.
Analysis and what to watch next
These reports highlight a broader trend in AI development toward product consolidation. By merging memories, Anthropic is moving away from siloed tools in favor of a “super app” architecture where context flows freely across different functional modes. This reduces the friction of rebriefing the AI but increases the permanence and persistence of user data within the Anthropic ecosystem.
For users, this means that their research preferences and project details will have a longer shelf life, potentially improving productivity for complex projects. However, it also places more importance on the accuracy of dynamic memory updates; if the AI incorrectly captures a detail during an active chat, that error will now propagate instantly to Cowork tasks.
Watch for how Anthropic manages “memory drift” over long-term interactions—where accumulated data might eventually conflict with new instructions. Additionally, users should monitor how often the system triggers notifications when recording sensitive information to determine if the balance between convenience and privacy remains functional in practice.