Anthropic has redesigned its language model Claude so that Chat and the task tool Cowork can now access a shared memory. Until now, users had to explain projects, preferences, and ongoing work again when switching between the two applications. Since August 25, 2026, this is no longer necessary: topics are saved during a conversation and are immediately available in the other tool as well.
Topics are now created live in conversation
Before the update, Claude summarized a conversation in a memory note only after it ended. The new system creates topics during the ongoing exchange, as Anthropic’s own release notes show. If users switch to Cowork, the tool for automated browser and office tasks, in the middle of a conversation, the assistant already knows the previous details.
Each saved point appears individually under the Topics section in the memory settings. From there, each entry can be read, modified, or completely removed without deleting the entire memory. Until now, Anthropic strictly separated notes from chat sessions and Cowork tasks, which often led users to explain context twice. According to TechCrunch, this repeated provision of information was considered one of the biggest friction points in the daily use of AI agents.
Users decide what remains in memory
Anthropic excludes certain information from memory by default: health data, ethnic origin, religious beliefs, political views, and gender identity are only stored by Claude if users actively enable the setting “Store sensitive topics in memory.” Identification numbers, social security numbers, criminal records, immigration status, and information that violates the terms of use are generally never stored in the system.
Those who want to keep Chat and Cowork deliberately separate can completely disable the shared memory, chat in incognito mode, or manually remove individual topics. According to The Register, there is no longer an option to exclude just one of the two tools from the shared memory – those who need strict separation must rely on separate accounts. The shared memory is also the basis for the usage dashboard Claude Reflect, which is only available for accounts with enabled memory.
Rollout distinguishes between personal and business accounts
The feature is now automatically enabled for Free, Pro, and Max accounts – free in the free plan, for around 18 euros per month for Pro, and from around 90 euros per month for Max. Team and enterprise organizations, on the other hand, start with disabled memory; administrators must first enable the option for their staff. Access is available immediately via web browser and desktop app, while mobile users on iOS and Android initially need to update to the current app version.
Anthropic does not explicitly mention a separate release for Germany or the EU; the feature follows the regular global rollout of the existing account plans. For companies with strict data protection regulations, it remains unclear whether and when team and enterprise customers will receive the shared memory by default – Anthropic has not yet provided a timeline for this.
It will be crucial whether Anthropic also enables the shared memory for team and enterprise customers, where data protection requirements are stricter and context must be shared among colleagues anyway. Competing products like ChatGPT and Microsoft Copilot are already building similar cross-account storage, so long-term memory functions could develop into a competitive field among AI assistants. Anthropic has not yet provided a date for the expansion to business customers.
