Claude Memory Now Unified Across Chat and Cowork

Introduction

On August 25, 2026, Anthropic announced that Claude's memory now works the same way in chat and in Claude Cowork. The update covers four changes: one shared memory pool across surfaces, real-time capture during conversations, user-editable memory files, and a new opt-in setting for sensitive topics.

Feature Overview

One Memory Shared Across Chat and Cowork

Anthropic's blog post states plainly: "Starting today, the memory you use in chat is the same as in Claude Cowork." When Cowork runs a task in the cloud, whatever Claude remembers from prior chats is available there, and information from a Cowork session is likewise available back in chat. Anthropic describes the change as "Cowork now has memory, and it's the same one you use in chat" — the capability arrives in Cowork and arrives already shared with chat.

Real-Time Capture Instead of Post-Conversation Summaries

Anthropic also changed how memory gets written. Previously Claude summarized conversations after they ended. Now, "Claude now adds topics to memory as you chat, instead of summarizing conversations after they end." Users retain the ability to pause memory or reset it at any time.

Editable, Transparent Memory Files

Every memory Claude holds is exposed to the user. "Everything Claude remembers is in a list of files under Topics in Memory settings, where you can read, edit, or delete each one." Anthropic notes the files are short, so correcting one file corrects what Claude knows everywhere that memory is used, rather than requiring the user to hunt down every downstream reference.

Sensitive Topics Are Opt-In

By default, Claude does not store personal or sensitive subject matter, defined by Anthropic as "your health, race, ethnicity, religious beliefs, politics, gender identity, and other similar areas." Users can turn on an "include sensitive topics in memory" setting; once enabled, a notice appears every time Claude saves a sensitive topic. The setting is not retroactive — "anything from before you turned it on isn't saved retroactively" — and it can be turned off again at any time.

Even with the sensitive-topics setting enabled, some categories remain hard-excluded: "sensitive identification numbers (SSN, government ID numbers, etc), criminal history, immigration status, or anything that violates" Anthropic's Acceptable Use Policy. Claude tells the user directly when it is unable to store something in that category.

Availability by Plan

Memory is on by default on Free, Pro, and Max plans across web, desktop, and mobile, though saving sensitive topics stays off by default even there. iOS and Android users need the latest app version to get the update. Team and Enterprise plans work differently: admins control availability for the organization, and memory stays off for individual users until an admin turns it on. All of the controls live under Settings > Memory.

Background

Claude's memory feature originally launched on September 11, 2025 for Team and Enterprise plans, with project-scoped memory and an Incognito chat mode. It expanded to Pro and Max plans on October 23, 2025. The August 25, 2026 update is the first time memory has been unified across chat and Cowork specifically.

Usability Analysis

The practical effect of unification is that a user no longer has to re-explain context to Cowork that they already established in chat, or vice versa. For anyone running Cowork tasks that build on ongoing conversations, this removes a friction point that previously required manually repeating preferences or project details in each surface.

The shift from post-conversation summarization to continuous capture changes the user experience in a subtler way: memory now updates while a conversation is still happening rather than only after it ends. Combined with the editable Topics file list, users get more visibility into what Claude has stored, and can correct it immediately rather than waiting for a summary pass. Anthropic's design intent is that a single edit to a memory file propagates everywhere that file is used, which is a meaningfully different model from having separate memory stores per surface.

The sensitive-topics opt-in is the most consequential privacy control in this release. Because it defaults to off and requires an explicit toggle with a per-save notice, users retain control over whether Claude retains information about health, politics, religion, or similar categories — while hard exclusions like SSNs and criminal history remain off-limits regardless of the setting.

Pros and Cons

Pros

1. Removes the need to re-explain context between chat and Cowork, since both now draw on the same memory
2. Real-time topic capture means memory updates as a conversation happens, rather than only in a post-conversation summary
3. Memory is fully user-editable through Topics files in settings, with a single edit propagating everywhere
4. Sensitive topics remain excluded by default, with an explicit opt-in and a per-save notice when enabled
5. Hard-excludes identification numbers, criminal history, and immigration status even when the sensitive-topics setting is on

Cons

1. Memory being on by default for Free, Pro, and Max is a real change in privacy posture compared to the original 2025 opt-in-only rollout
2. Continuous during-conversation capture means more can be written to memory before a user has a chance to review it
3. The list of excluded sensitive categories is defined entirely by Anthropic, and the post does not quantify how accurately that classification works in practice
4. Team and Enterprise plans default to memory being off, admin-gated behavior that is inconsistent with the Free/Pro/Max default

Outlook

Unifying memory across chat and Cowork is a step toward treating Claude as a single continuous assistant rather than a set of separate surfaces with separate context. The editable Topics files give Anthropic a way to keep that expansion user-auditable, which will matter more as memory scope grows. The sensitive-topics opt-in, with its per-save notice and hard exclusions, sets a specific precedent for how Anthropic is choosing to handle personal-category data as memory becomes more persistent and more automatic. How this plays out for Team and Enterprise customers, who get a different admin-gated default, will be worth watching as organizations decide whether to turn memory on for their users.

Conclusion

This update makes Claude's memory consistent across chat and Cowork, switches capture from post-conversation summaries to real-time topic tracking, and keeps memory fully visible and editable through Topics files in Settings. The sensitive-topics feature stays off by default with explicit opt-in and hard exclusions for the most sensitive categories. The update is most relevant to existing Claude users who move between chat and Cowork regularly, and to Team/Enterprise admins who will need to decide when to enable memory for their organizations.

Editor's Verdict

Claude Memory Now Unified Across Chat and Cowork earns a solid recommendation within the Claude space.

The strongest case for paying attention: removes the need to re-explain context between chat and Cowork, since both now draw on the same memory. That alone raises the bar for what readers should expect in this space. Reinforcing that, real-time topic capture means memory updates as a conversation happens, rather than only in a post-conversation summary — practical value rather than just headline appeal. The broader signal worth registering is straightforward: unifying memory across chat and Cowork removes the need to re-establish context when switching between the two surfaces. On the other side of the ledger, one constraint is real rather than a marketing footnote: memory being on by default for Free, Pro, and Max is a real change in privacy posture compared to the original 2025 opt-in-only rollout. It should factor into any serious decision. Layered on top of that, continuous during-conversation capture means more can be written to memory before a user has a chance to review it — which narrows the set of teams for whom this is an obvious yes.

For Anthropic and Claude users, alignment-focused teams, and developers already invested in the Claude ecosystem, this is a serious evaluation candidate, not just a curiosity to bookmark. For everyone else, the safer posture is to monitor coverage and revisit once the use cases that matter to your team are demonstrated in the wild.