Claude Cowork finally remembers what you told the app in chat

Anthropic has revamped Claude’s memory system, addressing one of the most aggravating aspects of working with AI agents. Prior to this change, users had to repeatedly reteach Claude information it already knew, depending on whether the interaction was happening in chat or Claude Cowork. Now, Claude will consistently recall what it learned in one interface, regardless of the platform.

Furthermore, users can now view, edit, or delete information Claude has retained in memory. This enhancement bridges the research and action phases more smoothly, preventing the need to start over when implementing ideas. Previously, users would need to reiterate details for projects they had brainstormed over time within Claude before using Cowork to take action.

The updated system ensures Claude operates as a seamless single assistant, avoiding the disjointed experience of two separate products. For instance, when drafting an update for a manager or conference agenda, Cowork would already be aware of relevant details such as the headcount, city, and speakers, thanks to past conversations.

Claude will now also add new topics to its memory during chats, rather than summarizing conversations at the end, making it easier to transition between chat and Cowork. Anthropic specifies that Claude won’t store personal or sensitive data by default, including health information, race, ethnicity, religious beliefs, political affiliations, gender identity, among others.

Users can, however, enable the storage of sensitive topics by turning on the “include sensitive topics in memory” option. Claude will also alert users whenever sensitive data is saved. Notably, the memory feature is already active on Free, Pro, and Max plans across web, desktop, and mobile platforms. On iOS and Android, users need to update to the latest app version to take advantage of this improvement.