What changed
Claude chat and Claude Cowork now read from and write to the same memory. Cowork picks up context from earlier chats: a project, a client, a preference you stated once. Anything Cowork learns during a task carries back into chat too.
Memory also updates live now. Claude used to summarize a conversation after it ended. Now it adds topics as the conversation happens, so a decision made mid session is already available if you switch tools before you’re done talking.
What did not change
Claude Code keeps its own memory, separate from this merge. If your workflow includes Claude Code for actual development work, don’t expect this update to touch that context. Chat and Cowork are unified. Code stands apart.
Also worth flagging for anyone running Cowork locally: the shared memory only applies when Cowork runs in the cloud. Local sessions don’t draw on it.
Why it matters for solo builders
On Canopy Guard, my security audit tool, I review findings in chat, then switch to Cowork to draft the client deliverable or the fix plan. Before this update, that switch cost me a rebrief every time. Now Cowork already has the site, the score, the finding.
If you run lean and move constantly between reasoning and building, this closes a gap you were probably absorbing without naming it.
Transparency and control
Anthropic exposed the memory itself. Settings shows stored topics, each one readable, editable, deletable. Sensitive categories, health, politics, gender identity, religion, stay off by default. You can turn them on if you want that level of personalization, and Claude flags it when a sensitive topic gets saved.
Rolled out by default on Free, Pro, and Max, across web, desktop, and mobile. Update your app if you haven’t already.
Top comments (0)