Anthropic has merged the memory systems of its Claude chatbot and its workplace assistant Cowork, meaning information Claude learns about a user is now automatically available to Cowork, and vice versa. The change, announced on Tuesday, is intended to make Cowork more useful for everyday office tasks, such as drafting a manager update or building event budgets, by drawing on context already established through chats with Claude, without users having to repeat themselves.

Anthropic says memories are now generated during conversations rather than only reviewed afterwards, and are switched on by default for Free, Pro and Max users, though Cowork itself is not available on Free accounts. "Sensitive" information, covering health, ethnicity, religion, politics and gender identity, is excluded unless users opt in, but once stored it is shared across both services. There is no way to keep Claude Chat and Cowork memories separate short of using different accounts; users can otherwise pause memory entirely, use incognito chats, or prune specific memories. Claude Code's memory remains separate, and Anthropic has not said whether that will change. Team and Enterprise admins control whether the feature is enabled for their organisations.