Claude Now Keeps One Memory Across Chat and Cowork, and Lets You Import What a Rival AI Already Knows About You, Logiciel Tech News

Claude Now Keeps One Memory Across Chat and Cowork, and Lets You Import What a Rival AI Already Knows About You.

Anthropic gave Claude a single memory that follows you across both chat and Cowork, so the assistant stops forgetting who you are every time you switch modes. The update went live on August 25. Hand Cowork a task and it now starts from what you already told Claude in chat: the project you talked through, your manager's preferences, the client from last quarter. As Anthropic put it, "When Cowork runs a task in the cloud, what Claude remembers from your chats is there, and vice versa." It's on by default for Free, Pro, and Max users on web, desktop, and mobile.

Robert

Logiciel Tech News

Aug 27, 2026 2 min read

Claude Now Keeps One Memory Across Chat and Cowork, and Lets You Import What a Rival AI Already Knows About You

What Changed

Until now, chat and Cowork kept separate memories, which meant re-explaining the same context depending on where you were working. That wall is gone. The two share one memory, and it updates as you talk instead of getting summarized at the end of a conversation. So if you mention halfway through a chat that a launch moved to October, Cowork knows that the next time it runs something for you.

This is the kind of fix that sounds small and isn't. The whole promise of an assistant that acts on your behalf falls apart if it can't remember what you said five minutes ago in a different window.

You Decide What's In It

The part Anthropic is leaning on is control. You can open the memory, read exactly what Claude has saved, and edit or delete any of it. Everything is sorted into named topics, so the entries look like a tidy list ("Meal Planning," "Client Pitch," "Half Marathon Training") rather than a black box. There are switches to stop Claude from referencing past chats or from generating memory at all.

Sensitive Things Stay Out Unless You Say So

By default, Claude will not save the categories most people don't want sitting in a profile: health conditions, race and ethnicity, religious beliefs, politics, and gender identity. If those happen to be the things you're tired of repeating, there's a toggle to let them in, and Claude notifies you when it saves one. A harder line exists for a few types of data it won't store at all, opt-in or not, including government IDs, Social Security numbers, criminal history, and immigration status.

That's a sharper default than some rivals ship with, and it's a reasonable one. Convenience and a permanent record of your health history are not the same trade for everyone.

It'll Even Pull From Your Old AI

The memory settings include an option to import context from another AI provider. Claude hands you a prompt to run in your other account, then pulls that history over so you don't start from zero. It's a small feature with an obvious target: people thinking about switching from ChatGPT or Gemini who don't want to abandon everything those tools have learned about them. Lowering the cost of moving is a quiet but pointed move.

Whether people actually trust a persistent, cross-surface memory enough to leave it on is the real test. The controls are there. Now it comes down to how many users bother to look at them.