Claude Now Remembers What You Tell It, No Matter Where You're Chatting

Anthropic is making Claude's memory work across both of its main tools, Claude chat and Claude Cowork. Now, if you tell Claude something in one place, it remembers that information when you switch to the other. The company announced the change on August 25, 2026, along with controls that let you see, edit, and delete what Claude has stored. (TechCrunch)

The way Claude captures information is also changing. Instead of waiting until a conversation ends and then writing a summary of it, Claude now saves individual topics as you talk. Think of it like taking notes throughout a meeting rather than writing one recap at the end. This means Claude is building up its memory of you during the conversation itself.

You can now see exactly what Claude has remembered. Anthropic lets you read, edit, or delete any piece of information the assistant has saved about a topic. This fixes a problem that has existed since AI assistants started keeping long-term memory: users had no way to check or correct what the model had decided to hold onto.

The privacy settings work in layers. By default, Claude will not store personal or sensitive information, including health data, race, ethnicity, religious beliefs, politics, or gender identity. If you want Claude to remember details in those categories, you can turn on an "include sensitive topics in memory" switch. Anthropic also says Claude will never save government-issued IDs, Social Security numbers, criminal history, immigration status, or anything else that violates its acceptable use policy, no matter what you choose.

The feature is widely available. Memory is turned on by default across Free, Pro, and Max plans on web, desktop, and mobile. iOS and Android users need to update to the latest app version.

This unification matters because it removes a wall that used to limit how useful Claude was. If you told Claude your project details or preferences in a chat session, that information would not carry over when you switched to Cowork, and vice versa. Now the assistant works more like one continuous tool no matter where you start talking to it.

The broader context here is about how companies handle sensitive data in AI memory. Anthropic's decision to leave health, political, and identity-related information out by default reflects caution about data types that carry higher legal and reputational risk. The toggle gives users control over their own boundaries, but the hard lines around IDs, Social Security numbers, and criminal history suggest the company has decided that storing certain data is not worth the compliance risk, even if the user agrees to it.

The switch from end-of-conversation summaries to saving topics as you go is a meaningful design choice. A summary at the end is easy to manage but may leave out details you cared about. Saving topics during the conversation preserves more specific information, but it raises a different concern: you may not know which statements Claude decided to keep until you check the memory view yourself. That is why the ability to see what Claude has stored is not just a nice feature but a necessary one.

Persistent memory is an area where many assistant products are actively competing. The tension between remembering useful context and giving users control over what gets kept is not unique to Anthropic. But the level of detail in these controls, especially the per-category sensitivity toggle and the hard exclusions on specific data types, sets a relatively high standard for managing that tension.

For developers and teams using Claude, the shared memory changes how multi-surface workflows operate. A preference stated in a chat session, a coding rule established in Cowork, or a project constraint mentioned in either context now carries over to both. Teams will want to think about how they structure interactions, especially around shared accounts where one person's memory could mix with another's.

In this author's view, the decision to turn memory on by default rather than asking users to opt in prioritizes a seamless experience over collecting as little data as possible. Users who want a fresh start each time they talk to Claude will need to manage or disable the feature themselves. Anthropic is betting that the convenience of having Claude remember things outweighs the effort of managing that memory, and for most users that is probably the right call. The ability to view and edit what is stored is the safety valve that makes turning it on by default reasonable.