Anthropic merges Claude's memory across chat and Cowork
Anthropic has unified the memory systems of Claude chat and Cowork, so users no longer need to repeat project context between the two. The AI now saves topics during conversations and lets users review, edit, or delete stored information. The feature is enabled by default on all plans, with sensitive data excluded unless users opt in.
The unified memory system now operates across web, desktop, and mobile platforms, with iOS and Android users required to update their apps to access the feature. Anthropic has built in several safeguards: sensitive categories like health data, political views, and religious beliefs remain excluded unless users explicitly toggle them on, while government IDs, Social Security numbers, and immigration status are never stored under any circumstances.
The technical shift also changes how memories are captured. Rather than summarizing a conversation only when it concludes, Claude now logs topics continuously during the exchange, allowing faster synchronization between chat and Cowork even mid-conversation. This addresses a longstanding friction point where users had to manually re-enter project details, such as conference headcounts or speaker lists, when transitioning between research and execution phases.
This update could meaningfully reduce workflow friction for professionals who rely on AI assistants across planning and execution stages, potentially accelerating project timelines. However, persistent memory also raises privacy considerations—users may forget what the AI retains, and the opt-in mechanism for sensitive topics places responsibility on individuals to manage their own data boundaries. The transparency features (review, edit, delete) may help mitigate this, but habitual users could still benefit from periodic memory audits to ensure stored information remains accurate and appropriate.