Anthropic has connected the memory of its two Claude products. Until this week, Claude the chatbot and Claude Cowork, the tool that opens your actual files and folders to complete tasks, kept separate memories. You could spend weeks planning something in chat, then have to explain the whole thing again the moment you asked Cowork to actually do the work.
That gap is now closed. Whatever Claude picks up in a chat conversation is available the moment you switch to Cowork, and the reverse is also true. Anthropic's own example is a conference agenda: if you already discussed the headcount, city, and speaker list in chat, Cowork will pull that context in automatically rather than asking you to re-type it.
This matters more than it sounds. Cowork launched in January as a way for non-technical people to hand off real work, things like sorting a folder of receipts into an expense report or turning a stack of notes into a finished document. Anthropic has been building it out fast since then, adding cloud access, mobile support, and in April a full business release with spending limits and admin controls for companies. A tool built to save people time was undercut by its own memory gap. Every switch between planning and doing meant retyping context, which is exactly the kind of friction that makes people give up on a new tool after the first try.
There's also a control layer worth noting. Anthropic is letting users see exactly what topics Claude has stored, and edit or delete them directly. By default, sensitive categories like health information, religion, political views, and gender identity are left out of memory entirely. Users can turn that off if they want Claude to remember those too, and they'll get a notification each time something sensitive gets saved. Certain data, like government ID numbers, criminal records, and immigration status, stays out no matter what setting is chosen.
That control matters more than it looks, because it's landing at a moment when businesses are genuinely nervous about what employees paste into AI tools. Security researchers have already flagged a sharp rise in staff sending sensitive company data into AI assistants, often through personal accounts nobody at the company can see. Anthropic giving people a visible, editable log of what an AI remembers about them is a reasonable answer to that worry, even if it depends on people actually checking it.
Competitively, this is Anthropic catching up more than leaping ahead. ChatGPT has offered a running memory of your preferences and history for a while now, and it was arguably a bigger reason people stuck with it day to day than any single new feature. Anthropic had kept Claude's memory more compartmentalized, partly by design, and this update closes that gap rather than opening a new one.
For a business owner, the practical takeaway is simple. If you or your team already use Claude to plan things, this removes a real piece of daily friction and makes handing off the follow-through to Cowork more worth trying. If you don't use it yet, this is a decent moment to test it on something small and repetitive, like sorting invoices or drafting a recurring report, and see whether an AI that remembers your context actually saves the time it promises.