Claude Memory Now Connects Chat and Cowork
Anthropic is giving Claude one shared memory across chat and Cowork, with a new Topics view for reviewing and editing what it remembers.
Claude memory now works across chat and Claude Cowork, so information learned in one experience can carry into the other. Anthropic announced the change on August 25, making Cowork less dependent on a separate briefing when a task grows out of an earlier conversation.
The shared system is designed around continuity. A project discussed in chat can carry details such as priorities, project status, or preferences into a Cowork task, while information that Cowork learns during that work can become available back in chat. Anthropic's announcement gives examples including manager updates, conference planning, and recurring business reviews.
One Memory Across Chat and Cowork
Cowork previously had to work from the context available when a task was handed to it. The new setup gives chat and Cowork access to the same memory, which reduces the need to repeat details before an agent starts working.
That change also fits the way Anthropic has been positioning Cowork. The company previously expanded Claude Cowork to the web and mobile, showing that the product is being used for more than coding, while Record a Skill in Claude Cowork added a way to capture repeatable workflows.
Claude Memory Updates While Conversations Continue
Claude can now add topics to memory as a conversation continues rather than waiting for the conversation to finish. A change mentioned in chat can therefore be available to a later Cowork task sooner.
Anthropic also exposes those memories instead of keeping them hidden. The company says saved information appears as individual topic files under Settings > Memory, where each item can be read, edited, or deleted.
Sensitive Topics Stay Off by Default
Claude does not save sensitive subjects such as health, race, ethnicity, religious beliefs, politics, or gender identity by default. Users can enable the "include sensitive topics in memory" setting, and Anthropic says Claude will show a notice when it saves information in one of those areas.
Some information remains excluded even when that setting is enabled. Anthropic says this includes government identification numbers, criminal history, immigration status, and information that violates its Acceptable Use Policy.
Availability Across Claude Apps
Memory is enabled by default for Free, Pro, and Max plans across web, desktop, and mobile. iOS and Android users need the latest version of the Claude app, while Team and Enterprise administrators control availability for their organizations.
The update gives Cowork a more direct connection to the conversations that often lead to its tasks. That matters most when the work depends on details built over time, because the handoff from discussion to execution no longer starts with a blank slate.
Anthropic has also been building Cowork into a broader AI work environment, including the earlier Claude Cowork coverage on business operations. For comparison, Saganote's ChatGPT Work and Claude Cowork comparison looks at how the two products approach AI-powered workspaces.