Anthropic Merges Claude Chat and Cowork Memory for Continuous Context
Anthropic has updated Claude’s memory infrastructure, combining the memory systems behind its chat interface and Claude Cowork to retain shared context across different modes of interaction. Under the new system, Claude remembers details gathered during earlier conversations when completing tasks in Cowork, removing the requirement to repeatedly brief the assistant on previously shared information.
What Happened
According to an announcement made by Anthropic on Tuesday, Claude now operates on a merged memory foundation spanning both chat and Claude Cowork. Previously, users who brainstormed or researched projects in standard chat sessions had to manually re-explain the specifics when moving into Cowork to carry out actions.
With this change, Claude connects the two environments so information documented in one space carries into the other. Furthermore, the system now logs topics in real time while a chat is underway, rather than waiting for the dialogue to conclude before building a summary. This real-time logging allows information to transfer immediately into Cowork, even while a user continues an ongoing conversation.
Anthropic has also added user controls that display everything Claude stores in its memory. Users can review, adjust, or completely remove stored entries across any topic.
Key Highlights
- Unified Knowledge: Claude retains information across both chat and Cowork, treating interactions as a continuous assistant rather than segmented tools.
- Real-Time Storage: Memory updates dynamically as conversations progress instead of processing solely at the end of a chat session.
- User Transparency and Control: Individuals can view, edit, or delete any facts or topics preserved in Claude’s memory database.
- Privacy Defaults: Information concerning health, political opinions, religious beliefs, race, ethnicity, and gender identity is excluded from memory by default. Users can optionally toggle on an option to include sensitive topics, which triggers a notification whenever such data is logged.
- Permanent Exclusions: Claude will strictly not store government identifiers, Social Security numbers, criminal records, immigration status, or items that breach Anthropic’s acceptable use policy.
- Broad Platform Rollout: The memory capability is turned on by default across Free, Pro, and Max tiers on web, desktop, and mobile platforms.
Why This Matters
Working with AI tools has often involved friction when transitioning between initial planning and final execution. For instance, if an individual discusses conference planning details in chat—such as venue cities, attendee headcounts, and guest speakers—they would previously have to re-enter those exact details when asking Cowork to construct a conference agenda or an update for a manager. By connecting the two environments, Claude preserves the context established during the research stage, making it functional during task execution.
What to Watch Next
The shared memory functionality is accessible across web, desktop, and mobile applications. To utilize the integrated memory on smartphones, users on iOS and Android must install the latest version of the Claude app.
Frequently Asked Questions
Which subscription tiers have access to the unified memory?
Anthropic has enabled the memory feature by default across Free, Pro, and Max plans on web, desktop, and mobile platforms.
Can users manage what Claude remembers?
Yes. Anthropic allows users to view the information Claude has saved and offers the ability to edit or delete stored memories on any topic.
Does Claude store sensitive or identification data?
By default, Claude does not save sensitive personal categories like medical details, religion, politics, race, or gender identity, though users can manually opt in to include sensitive topics. The system will never store government IDs, Social Security numbers, immigration status, or criminal history under any circumstances.
Source: Anthropic via TechCrunch