Claude's Memory Now Syncs Across Chat and Cowork

Anthropic fixed one of the more tedious parts of using Claude on Tuesday, announcing that the chatbot's memory system will now carry over between its chat interface and Claude Cowork, its agentic work tool. Until now, the two ran on separate memory stores, so anything you told Claude in a regular conversation stayed there. Ask Cowork to actually act on those plans, and it had no idea what you were talking about. You'd end up repeating headcounts, city names, deadlines, whatever context you'd already spent time building, just to get the agent moving.

That gap made Claude feel less like one assistant and more like two products sharing a login. Anthropic's fix merges the memory layers so information learned in one surface is available in the other. The company is also giving users direct access to what Claude has stored, with the ability to view, edit, or delete anything by topic. There's a smaller shift too: Claude now logs details into memory as a conversation happens, rather than waiting until the chat ends to summarize it.