Anthropic gives chat and Cowork one memory
The New Stack Frederic Lardinois ● Covered by 3 sources
Anthropic tied Claude’s chat and Cowork memory together. Now it can carry context across both, so you don’t keep repeating yourself.
Based on reporting by The New Stack, Frederic Lardinois — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Anthropic rolled out a major change to Claude on Tuesday: the assistant’s memory for chat and Cowork is now one system instead of two separate ones. Before this, Cowork memory lived inside individual projects and didn’t draw on what Claude had learned about you in chat. Now, Anthropic says, Claude should remember you no matter where you use it.
That matters most in Cowork, where people are more likely to spend hours or days on deeper work. Anthropic’s pitch is simple: fewer repeat explanations, more continuity. If Claude already knows who your manager is, or that you’re planning a conference, it can carry those details into follow-up work instead of making you spell them out again.
The memory system also updates while you’re still working. Earlier, those updates happened only after a chat ended. Anthropic says that now the context built up over months of conversations — things like priorities and project status — is available the moment you hand Cowork a task, and can carry back into chat too.
There are limits. Claude won’t, by default, store sensitive material such as health data or information about religion, race, ethnicity, or gender identity. Users can choose to allow those topics in settings, but Anthropic doesn’t offer a finer-grained middle ground; it’s all or nothing. Some information is blocked entirely, including sensitive identification numbers, criminal history, immigration status, and anything that violates Anthropic’s Acceptable Use Policy.
The memory files themselves are small topic-based files. You can delete them, but not edit them directly. If you want changes, you can chat with Claude about them from the settings menu. The unified memory feature is on by default for Free, Pro, and Max users on web, desktop, and mobile, while Enterprise and Teams accounts have memory features off by default.
My take — AI-written commentary, not fact-checked reporting
Anthropic is making the right move here. Memory is only useful if it follows the user across the places where work actually happens, not trapped in little product silos like a bureaucrat with a filing cabinet. The ad-free jab at OpenAI is doing some heavy lifting, but the real point is simpler: useful memory beats creepy memory, and Claude is trying to sound like it understands that.
Read more about this at: The New Stack
