Claude's memory now works the same in chat and Cowork

Anthropic merged Claude’s memory system across chat and Cowork on August 25, so the same stored context now follows a conversation into a Cowork task and back. Previously the two ran separate memory layers, meaning a project detail you’d explained in chat had to be re-explained once you handed a task to Cowork, Anthropic’s agent for longer, cloud-run work. Memory now updates live as you chat rather than only after a conversation ends, so it becomes available to Cowork faster.

It’s already switched on for anyone on Free, Pro, or Max, on the web, desktop, and mobile apps alike; a Team or Enterprise admin has to flip it on for their organization. Saved memories are organized by topic in Settings, where users can read, edit, or delete any of them, and pause or reset the whole thing at any time. Health conditions, race, ethnicity, religious beliefs, politics, and gender identity are excluded by default even when Memory is on, and Claude flags it when it saves something in a sensitive category; users can opt in to storing those topics if they choose.

What it means for you

If you run recurring tasks through Claude Cowork, this closes a real gap: a preference or constraint you stated once in chat, like a formatting rule or a system you always want excluded, should now carry over automatically instead of needing to be pasted into every task brief. The tradeoff is the same one covered in our team memory and privacy read: more shared context means more surface area for one bad instruction, or one piece of client information mentioned offhand, to persist and resurface somewhere you didn’t expect. Worth a five-minute audit of Settings > Memory before you lean on this, especially on a Team or Enterprise seat where an admin controls the default for everyone.