Claude Memory Now Works Across Both Chat and Cowork Sessions

Anthropic has rolled out a major update that lets Claude remember things across both its chat feature and Cowork, its agentic work tool. Before this change, memory in these two spaces stayed separate. Now, details shared in a normal conversation can carry over when a task is handed to Cowork, and the reverse is also true. This update matters because it removes a common frustration: having to repeat the same background information every time you switch between talking to Claude and asking it to complete a task.

The update was announced on Tuesday, August 25, 2026. Anthropic said that when Cowork runs a task in the cloud, information Claude remembers from past chats will be available right away. The same works the other way around, so anything Cowork learns while working on a task can show up later in a normal chat. This means users no longer need to explain the same project details, like deadlines or team preferences, more than once.

Cowork is Anthropic’s computer-use agent, built to handle repetitive or multi-step tasks like organizing files, building documents, or working across different apps. Until now, its memory was tied to individual projects and had no connection to what Claude knew from chat conversations. Anthropic has been working to bring chat and Cowork closer together for some time, and this update is described as a major step in that direction.

Another change is how Claude updates its memory. In the past, Claude would only create a summary of what to remember after a conversation ended. Now, it can add new information to memory while the conversation is still happening. This is especially useful for Cowork, since those sessions can run for hours or even days, so waiting until the end to save memories would not make sense.

Anthropic is also giving users more control over what gets stored. By default, Claude will not save sensitive details, such as information about a person’s health, race, ethnicity, religious beliefs, or gender identity. However, some users may want this kind of information remembered, so Anthropic has added a setting that lets people choose to include sensitive topics if they wish.

Users can view what Claude has stored in its memory, organized by topic, and can edit or delete any of it through the settings menu. Anthropic also said certain highly sensitive items, like government ID numbers or immigration status, will never be stored at all. The memory update works by default on Free, Pro, and Max plans across Claude’s web, desktop, and mobile apps, though mobile users need to update to the latest version to get it. Team and Enterprise users can have it turned on by an account owner.

In short, Claude’s memory system now works as one shared system instead of two separate ones. This should make it easier for people to move between chatting with Claude and delegating work to Cowork without losing context. Anthropic frames the update as part of a bigger effort to make Claude function less like a simple chatbot and more like a continuous work assistant. Further updates to memory controls and agentic features are likely as this effort continues.