Anthropic is changing how memory works across Claude and Claude Cowork, aiming to remove a common friction point for people who use AI agents for both planning and execution.
Announced on Tuesday, the update merges the memory system used by chat with the one used by Claude Cowork. In practical terms, Claude can now retain what it learned in one part of the product and use that context when the user switches to another.
What Is Changing
The core update is simple: Claude chat and Claude Cowork will share memory. A user who develops a project idea in chat should not have to explain the same details again when moving into Cowork to draft, organize, or act on the work.
That matters because Claude Cowork is designed for action-oriented tasks. Before this change, a person could spend time building context in chat, then face a fresh start when asking Cowork to use that context. The result was a break between research and execution.
Anthropic is positioning the update as a way to make Claude feel less fragmented. Instead of treating chat and Cowork as separate experiences, the product can now carry relevant information across both.
Why Shared Memory Matters
For users, the most immediate benefit is less repetition. AI agents often ask for information that the user has already provided elsewhere in the same broader product. That can make the workflow feel slower than it should, especially when the task depends on accumulated context.
Anthropic gives examples such as drafting an update for a manager or preparing an agenda for a conference. If the relevant details were already discussed in prior conversations, Cowork could understand references without requiring the user to re-enter them.
In the conference example, Claude could already know details such as the headcount, city, and speakers. That means the user can move straight into the actual task instead of rebuilding the background first.
This is especially useful for work that develops over time. A person may refine plans, clarify goals, and add details during several chats. With shared memory, those pieces of context become more available when the user moves from thinking through an idea to producing something from it.
Memory Updates While You Chat
Anthropic is also changing when Claude adds topics to memory. Rather than waiting until a conversation ends and summarizing it afterward, Claude will now add topics as the chat continues.
That makes the shared system more current. If a user is still in the middle of a conversation, new information can become available across chat and Cowork more quickly than before.
The result is a smoother handoff between the two experiences. A user can explore an idea, adjust the details, and then shift into Cowork with fresher context already available.
The source article describes this as a practical improvement for moving between the research phase and taking action. The value is not only that Claude remembers more, but that the memory is updated in a way that better matches how people actually work.
User Controls And Sensitive Information
Anthropic is exposing what Claude has stored in memory. Users will be able to read, edit, or delete information on any topic.
Those controls are important because shared memory can be helpful only if users can see and manage what has been retained. If Claude is going to rely on remembered context, the user needs a way to correct it, remove it, or inspect it.
Anthropic says Claude will not store personal or sensitive information by default. The examples named include health data, race, ethnicity, religious beliefs, politics, gender identity, and more.
Users who want Claude to remember sensitive topics can enable that by turning on "include sensitive topics in memory". If a sensitive topic is saved, the app will notify the user.
There are also categories Claude will not save. Anthropic says this includes government-issued ID and Social Security numbers, criminal history, immigration status, and other information that violates its acceptable use policy.
Who Gets The Update
The memory feature is enabled by default for Free, Pro, and Max plans. It applies across web, desktop, and mobile.
For people using Claude on iOS and Android, there is one extra step: the app must be updated to the latest version.
The broader significance is that Claude Cowork can now operate with more of the context users have already built in chat. That does not change the basic need for users to review important outputs, but it does reduce the amount of repeated briefing needed to get useful work done.
For anyone moving between brainstorming, planning, and drafting inside Claude, the update is meant to make the assistant feel more consistent from one step to the next.