AI
Claude unifies chat and Cowork memory so AI remembers context across modes
Anthropic merged its memory system across Claude's chat and agent interfaces, eliminating the need to repeat information when switching between planning and execution.
TechCrunch — AI
Anthropic has addressed a long-standing friction point in AI assistance: the requirement to repeatedly brief an agent on details it already learned in a different conversation mode. By consolidating the memory systems that power Claude's chat interface and Claude Cowork (its autonomous agent platform), the company is creating what feels like a single, continuity-aware assistant rather than two disconnected tools.
What's changing#
The unified memory approach means information retained during chat sessions now automatically carries over to Cowork sessions—and vice versa. Users can now browse, edit, or delete stored memories directly, giving them transparency and control over what Claude retains.
Instead of waiting until a conversation ends to capture and summarize key details, Claude now updates its memory incrementally as users chat. This real-time approach keeps Cowork in sync with ongoing discussions, making it faster and smoother to switch between brainstorming and action.
Why this matters#
Previously, workflows often required restarting from scratch: you might develop ideas for a conference agenda through discussion with Claude, but when it came time to have Cowork draft the actual update or materials, you'd have to re-enter headcount, location, speaker names, and other context. The merged memory system eliminates that friction, letting users move fluidly from research and planning phases into execution without lost information.
Privacy controls#
By default, Claude avoids storing sensitive personal data—including health information, political beliefs, religious affiliation, gender identity, and ethnicity. Users who want to include such topics can opt in via a toggle, and the system will alert them whenever sensitive material is being saved.
The company maintains hard boundaries on certain data types that Claude will never store, regardless of user preference: government IDs, Social Security numbers, criminal history, immigration status, and content violating acceptable use policies all remain off-limits.
The feature rolls out by default across Free, Pro, and Max plans on web, desktop, and mobile. Mobile app users will need to update to the latest version to access it.

