Technology
Anthropic unifies Claude memory across chat and Cowork features
The update lets the AI retain user context across different tasks while adding controls to edit memories or store sensitive topics.
The short version
- Anthropic merged the memory systems for Claude chat and its Cowork agent, allowing context to carry over between conversations and cloud tasks.
- Claude now saves memory topics in real time as conversations happen, rather than creating summaries after a session ends.
- Users can view, edit, or delete stored memories, with an optional setting to allow sensitive topics like health and politics while excluding high-risk data like Social Security numbers.
Key facts
- Anthropic updated Claude to unify memory across standard chats and Claude Cowork agent sessions.[CNET · 9to5Mac · Engadget · TechCrunch]
- Claude updates its memory in real time during active chats rather than compiling a summary at the end of a conversation.[CNET · TechCrunch]
- Users can review, modify, or clear stored memories by topic through the settings menu.[CNET · Engadget · TechCrunch]
- Storage of sensitive categories such as health, religion, or politics is disabled by default, but users can explicitly opt in to save them.[CNET · 9to5Mac · Engadget · TechCrunch]
- High-risk data including Social Security numbers, government IDs, criminal records, and immigration status remain strictly excluded from memory even if sensitive topic storage is enabled.[CNET · Engadget · TechCrunch]
- The unified memory functionality is active by default across Free, Pro, and Max tiers on web, desktop, and updated mobile apps.[CNET · 9to5Mac · Engadget · TechCrunch]