What is Memory, and how is it different from the context window Claude already has?
The context window is the range of content Claude can read and reference within a single conversation — everything you've said in that same conversation, Claude can see, but once you start a new conversation, that content typically doesn't carry over automatically. Memory solves exactly this problem: it's a persistent memory that spans across conversations and across surfaces (shared between chat and Cowork). Something you mentioned in one conversation is available to Claude even in a brand-new conversation, on a different device, as long as you're signed into the same account.
Put simply, the context window governs "how much Claude remembers within this one conversation," while Memory governs "what Claude remembers about you across every conversation you've ever had" — the two are complementary, not a replacement for each other.
Why did Claude introduce Memory, and what problem does it solve?
The most direct problem is the cost of repeated explanation — if you have to re-explain "what project I'm working on," "what response style I like," and "who's responsible for what on my team" every time you start a new conversation, that repeated input adds up to real time loss, and it's easy to accidentally leave out a detail, causing Claude's answer to miss important context.
The deeper reason is that Claude's product line is moving toward treating chat and Cowork as two interfaces on the same underlying system. Previously these were separate experiences, and users had to mentally track "which interface knows what." Memory removes that burden — if you discuss a budget for an event in chat, and Cowork later builds the logistics document, it already knows the headcount, the city, and the speaker list, without you having to say it again.
What does the actual process of using Memory look like? Do I need to explicitly ask Claude to "remember" something?
No, you don't. Memory operates in real time as you're having the conversation — you don't need a specific command like "remember this." As long as something you mention gets judged as worth remembering (a project deadline, your preferred communication style, a team member's name), Claude adds it to Memory directly, without waiting until the conversation ends to organize it.
The actual place to view and manage it is Settings > Memory, where content is organized by Topics into a list you can read entry by entry. If you notice an entry got something wrong — an outdated company name, say — you can edit or delete that specific item directly, and the corrected version applies to every conversation going forward, without needing to re-explain the background. You can also pause Memory at any time, or reset it entirely — all from the same settings page.
I just started using Claude — what should I keep in mind about Memory as a beginner?
The first thing to check is your plan type. If you're on Free, Pro, or Max, Memory is on by default, meaning starting now, anything worth remembering that comes up in conversation gets saved automatically — if you don't want that, you can go into settings and pause or turn it off yourself. If you're using Claude through a company Team or Enterprise account, Memory is off by default, and whether it's available to you depends on whether your organization's admin has enabled it.
The second thing to note is the default behavior around sensitive topics: information about health, religion, or political views won't be saved to Memory by default, even if you mention it in conversation — you don't need to be especially careful about avoiding these topics. But if you want Claude to remember this kind of thing going forward (a food allergy affecting restaurant recommendations, for example), you need to go into settings and turn on the corresponding option yourself, rather than assuming it'll get saved just because you mentioned it once.
The third practical habit worth building is occasionally checking Settings > Memory to glance at the current list — especially if you notice Claude's answers seem to be assuming some outdated piece of background. That's often a sign a specific Memory entry needs correcting, and fixing it directly in the list is a lot less work than re-explaining it in every conversation.
If you've only recently started using Claude, you may have noticed something: in some conversations, Claude seems to already "know" a project's status or how you like to be addressed, without you re-explaining it. That's not a coincidence — it's Claude's Memory feature, newly launched in August 2026. Understanding how it works, what gets saved by default, and what doesn't is worth sorting out early as a beginner.
Previously, if you used Claude's regular chat and Claude Cowork separately, the two didn't share context with each other. Now they share the same Memory: project priorities or who's who on your team that you've mentioned in chat are directly available to Cowork when it runs a task, and context that comes up while Cowork handles a task feeds back into your later chat conversations too. That means you no longer need to re-explain background information in every interface — say it once, and it carries forward wherever you go next.
Another thing beginners commonly misunderstand is that Memory doesn't wait until a conversation ends to go back and "summarize" the key points. Instead, the moment you mention something worth remembering during a conversation, Claude adds it to Memory in real time. If you mention a project deadline moved to September, you don't need to say "remember this" — by your next conversation, Claude already knows.
Yes, you can. Everything Claude remembers is organized by topic into a list under Topics in Settings > Memory, where you can read, edit, or delete each entry. This means Memory isn't a black box — if Claude got something wrong (an outdated company name, for instance), you can correct that one entry directly in the list, and future conversations will use the corrected version without you having to re-explain from scratch.
The thing most worth understanding as a beginner is this: health conditions, race, religious beliefs, political views, gender identity, and similar sensitive topics are not saved to Memory by default, even if you mention them in conversation. If you want Claude to remember this kind of information — a food allergy, say, so it's factored into meal planning later — you need to go into Memory settings yourself and turn on "Include sensitive topics in memory." Once turned on, you'll see a clear notice every time Claude saves something in that category, and the setting only affects conversations going forward — sensitive content you mentioned before turning it on isn't retroactively saved.
Separately, regardless of whether that setting is on, there's a category of information Claude never saves to Memory under any circumstances: government ID numbers, criminal history, immigration status, or anything that violates Anthropic's Acceptable Use Policy. If you ask Claude to remember something in this category, it will tell you directly that the request can't be fulfilled, rather than silently skipping it.
Memory is currently on by default on the web, desktop, and mobile apps for Free, Pro, and Max plans. For Team and Enterprise organizations, it's off by default — an admin needs to enable it for the organization before individual users can choose to turn it on themselves. If you're on mobile, make sure you've updated to the latest version of the app to see the full Memory feature.