Claude's Memory Now Follows You From Chat Into Cowork

 Anthropic announced that Claude's memory now works the same way whether you're chatting or using Claude Cowork, and that every person using it can see exactly what's being remembered and change it. That's a meaningful shift from memory as a background feature to memory as something users actively manage. 

Key Points

• Claude's memory is now unified across chat and Claude Cowork, so context built in one carries into the other instead of requiring users to re-explain projects, preferences, or people each time they switch surfaces.
• Memory updates in real time as a conversation happens, rather than being generated as a summary after the conversation ends, so a detail like a moved deadline is available in the very next chat.
• Every saved memory is visible and editable under Topics in Memory settings, stored as short files a user can read, correct, or delete individually.
• Sensitive topics like health, religion, and political views are excluded from memory by default, but users can opt in to have them saved, with a notice shown each time a sensitive topic gets stored going forward.
• Memory is on by default for Free, Pro, and Max plans, while Team and Enterprise admins control availability and it stays off for individual users until they turn it on.

What Changes When Memory Works Across Chat and Cowork

Previously, context built up in a chat conversation didn't necessarily carry over when a task moved to Cowork, Anthropic's agentic work tool, or vice versa. According to Anthropic's announcement, that gap is now closed: context from months of chat conversations, like a team's Q3 priorities or a project's status, is available the moment a user hands a task to Cowork, and anything Cowork learns while completing a task carries back into chat. The company's own examples include asking Cowork to draft a manager update where it already knows who the manager is and how they prefer updates written, or building a conference budget document that already reflects headcount and speaker details established earlier in chat.

Memory Now Updates Live Instead of After the Fact

The announcement also describes a change in how memory gets written. Rather than summarizing a conversation once it ends, Claude now adds topics to memory as the conversation happens. Anthropic's example is a deadline moving to September: mention it once, and the next conversation already reflects the change without the user needing to explicitly say "remember this." Users retain the ability to pause or reset memory at any time.

Full Visibility Into What Claude Remembers

Every memory Claude has saved appears under Topics in Memory settings as an individual, editable file. Anthropic frames the benefit as a single correction propagating everywhere: fixing an outdated company name in one file means every future conversation uses the corrected version, rather than requiring the user to catch and correct the same error repeatedly across separate conversations.

How Sensitive Topics Are Handled by Default

Claude does not store information related to health, race, ethnicity, religious belief, political views, gender identity, or similar categories in memory by default. Users who want Claude to remember that kind of detail, Anthropic's example is a gluten allergy relevant to meal planning, can turn on an "include sensitive topics in memory" setting, at which point they'll see a notice each time something in that category gets saved. That setting only applies going forward; nothing from before it's enabled gets added retroactively. Even with the setting on, certain categories are excluded outright regardless of user preference: government ID numbers, criminal history, immigration status, and anything that would violate Anthropic's Acceptable Use Policy. Claude will inform the user directly when it's unable to save something falling into one of those excluded categories.

Availability Across Plans

Memory is on by default for Free, Pro, and Max plans across web, desktop, and mobile, with sensitive-topic storage off by default even for those users. For Team and Enterprise accounts, admins control whether memory is available at the organization level, and it remains off for individual users until they choose to turn it on. Anthropic notes that mobile users should update to the latest app version to get the current functionality.

For teams evaluating how AI tools fit into a broader growth strategy, a unified, user-editable memory layer is worth watching closely, since it changes how much context an AI assistant can carry across a workflow without repeated manual briefing. If your team is figuring out how to build that kind of persistent, well-governed context into your own AI-assisted processes, that's exactly the kind of setup work our AI marketing services team helps clients think through.