Anthropic has expanded Claude’s memory feature to enable shared context between chats and Cowork sessions. Claude, who has been able to remember some information from previous conversations since last summer, can now transfer this information to the company’s Cowork tool, which focuses on computer use. Similarly, the context created during Cowork can be used in conversations with Claude. In this way, it is aimed to reduce the need for users to re-explain their preferences or previously provided information when they switch to different Claude tools. With the same update, Anthropic also offers users more detailed controls regarding the inclusion of sensitive topics in memory.
Until now, Claude’s memory feature could only work in a fragmented manner for different reasons. Part of this was related to Anthropic’s conscious product choices, while part of it was related to the limitations of Claude’s technical architecture. The new regulation reduces this distinction, especially in the usage between Claude and Cowork. According to Anthropic’s statement, when Cowork is running a task in the cloud, the information Claude remembers from the conversations will be available, and vice versa. In this way, previously determined working preferences with Claude, for example, may not have to be re-explained when a recurring task is performed on Cowork.
Claude memory feature becomes partner with Cowork
Cowork, which Anthropic launched at the beginning of the year, focuses on automating repetitive computer tasks through Claude. Users can transfer routine transactions they perform regularly during the day to Claude through this tool. Collaboration of memory between two environments becomes more meaningful, especially in such tasks, because the necessary context during automation does not have to be recreated in each session. In addition, Anthropic recently updated the Claude for Chrome extension, making Cowork accessible via the browser add-on. Following this change, Claude conversations held in Chrome’s side panel are also included in the user’s usage history.
While the company expands the scope of memory, it also increases the user’s control over the recorded information. In recent weeks, Claude users have been able to see the information stored in memory arranged according to subject headings and can change the details under each heading from Claude’s settings menu. With the new update, the “include sensitive topics in memory” option is added. This setting is kept off by default. According to Anthropic, Claude does not memorize topics considered personal or sensitive, such as health, race, ethnicity, religious beliefs, political views, and gender identity, in standard settings.
When the user activates the sensitive topics option, Claude will be able to keep some information in these categories in his memory. Despite this, the system is not expected to record this information without the user’s knowledge; Anthropic states that a notification will be shown whenever a sensitive topic is added to memory. Turning on the option does not mean scanning past conversations retrospectively and transferring sensitive information to memory. Users will be able to turn off the feature at any time and will have the opportunity to edit and delete saved information. The company explains this preference on the grounds that it may be useful for Claude to remember some information that is considered sensitive for other users.
Allowing the recording of sensitive topics also does not mean that Claude can retain all kinds of personal information. Anthropic specifically states that some types of information will not be saved even if the user has turned on the relevant option. These include highly sensitive identification information such as a US Social Security number, criminal history and immigration status. Thus, the company draws a separate line between sensitive issues that are left to the user’s choice and categories of data that should not be included in the memory at all. Especially considering that Claude started to use shared memory in different applications and working environments, the options of seeing what data is kept and deleting it become one of the basic elements of user control.
Claude’s memory feature can be used on Free, Pro and Max accounts and is enabled by default. Those who want to access new options on the mobile side must install the latest version of Claude’s Android or iOS application. The convergence of context between chats, browser usage, and Cowork tasks can reduce the need to provide the same information over and over again, especially for those who regularly use Claude in their daily work processes. On the other hand, the expansion of the scope of common memory makes it even more necessary to easily view, edit and delete recorded information. While Anthropic’s new controls offer more choice in this regard, the practical value of the feature will depend on the balance users will establish between continuity and personal data control.
Join Channel