Claude's Memory Now Works Across Chat and Cowork — Here's How to Control Everything It Saves

On Tuesday, August 25, 2026, Anthropic removed one of the biggest friction points in the Claude experience. The official announcement: "the memory you use in chat is the same as in Claude Cowork." In plain terms, everything you have explained to Claude in conversations — your quarterly priorities, the state of your projects, how your team thinks — is present the moment you hand Cowork a cloud task, and whatever surfaces inside Cowork flows back into your chats. You no longer re-brief in either direction.

The update — published in an official post titled "Claude's memory works everywhere, and you decide what's in it" — arrives with a substantial control layer: view everything Claude remembers, topic by topic; edit or delete any of it; and explicit handling of so-called "sensitive topics" such as health and beliefs, which are not stored by default in the first place. This piece explains the details exactly as stated in the primary source, what it means for you in practice, and the limits you should understand before relying on the feature.

Source: Claude official account on X

Source: Claude Blog — Claude's memory works everywhere

What Exactly Changed? One Memory Instead of Two

Before this update, the Claude experience was split: chat knew certain things about you, while the Cowork agent — which we covered when it launched on phone and web in July — started its cloud tasks from a near-blank slate, forcing you to re-explain context every time you delegated work. After the update, one memory flows in both directions:

• Ask Cowork to draft an update for your manager? It "already knows who that is and how she likes updates written" — in the official post's own words.
• Brainstorm a conference agenda in chat? When Cowork builds the budget and logistics doc, it "knows the headcount, the city, and the speakers."
• Explain once how your team defines its metrics? Every quarterly business review deck Cowork builds afterwards uses that definition, "with no rebriefing."

There is a second change to the mechanism itself: Claude now adds topics to memory while you chat, instead of summarizing conversations after they end. The company's example: mention that your project deadline moved to September, and your next conversation already knows — without you saying "remember this."

Managing Memory: The "Topics" Files You Can Read, Edit, and Delete

Everything Claude remembers now appears as a list of short files under "Topics" in Memory settings (Settings, then Memory). For each file you have three powers: read it, edit it, or delete it. The practical idea is clever: one correction propagates everywhere — fix your company's old name in a single file, and every subsequent conversation gets it right.

You can also pause memory or reset it entirely at any time from the same settings — an important option for anyone using Claude across multiple clients or distinct contexts.

Privacy: What the Memory Skips by Default, and What It Never Stores

This is the most consequential part of the announcement for anyone wary of "remembering" features:

• Sensitive topics are not stored by default: health, race, ethnicity, religious beliefs, politics, gender identity — Claude does not put these in memory unless you explicitly enable "include sensitive topics in memory."
• If you enable it, every save is announced: each time Claude stores something on one of these topics you will see a notice, and the setting only applies going forward — nothing from your earlier conversations is retroactively saved.
• Some things are never stored, even with the setting on: sensitive identification numbers (SSN, government ID numbers), criminal history, and immigration status are flat-out excluded from memory, and Claude will tell you when it is unable to save something. Anything violating the Acceptable Use Policy is obviously out as well.

In other words: the default design leans cautious, and expansion is an explicit, gradual opt-in — not an open door.

Who Gets the Feature, and When

Per the official announcement:

• Free, Pro, and Max plans: memory is on by default across web, desktop, and mobile. On iOS and Android, update to the latest app version.
• Team and Enterprise plans: the feature is under the organization admin's control, and it is off for individual users until enabled — a relevant point for workplace teams in the region using Claude organizationally.
• Geographic availability: no regional restrictions are mentioned in the announcement; the feature is available globally under the same plans.

What This Means for You as a Professional

• Genuine time savings: the biggest hidden cost of AI agents has been "re-briefing" — explaining context for every task. With unified memory, Claude becomes a reference that accumulates understanding of your projects over weeks, whether you write in Arabic or English.
• The cloud agent becomes truly practical: Cowork's real value was capped by how much it knew about you; now you can hand it extended tasks (building decks, compiling reports, drafting recurring updates) and it "picks up where you left off."
• For regional teams: no language or regional barrier is stated — memory works with Arabic content the same as English, consistent with what we found in our Claude vs ChatGPT comparison for Arabic.
• Flip it to your advantage safely: review the Topics list weekly at first, delete anything you do not want persisted, and keep sensitive topics off unless you truly need them.

Quick Comparison: Unified Memory vs What Came Before

For those comparing platforms: both ChatGPT and Gemini have had conversation-level memory features for a while, but the distinctive addition here is unifying memory between the chat interface and a cloud agent inside one product, together with the explicit sensitive-topics layer — a combination we have not seen in this exact shape from competitors to date.

Honest Limitations: Know Before You Rely On It

• Memory is on by default for individual plans: if you want full privacy, you must go into settings and switch it off yourself — good intentions are the default, but the action is yours.
• Sensitive-topic saving is forward-only: anything not saved before you enabled the setting will not be retroactively recovered.
• Team and Enterprise users wait: availability depends on your organization's admin, so access may lag.
• Memory is not a backup: these are short context files that serve personalization — not a complete archive of your conversations. Do not treat it as a vault for critical information, whatever the plan.

Frequently Asked Questions About Claude's Unified Memory

Is Claude's memory in chat and Cowork really unified now?
Yes, as of August 25, 2026: one memory flows between both interfaces in both directions, according to Anthropic's official announcement.

How do I see what Claude remembers about me?
Go to Settings, then Memory: the Topics list shows everything stored, and you can edit or delete any file, or pause or reset memory entirely.

Does Claude store my health or religious information?
Not by default. Sensitive topics (health, race, religion, politics, gender identity) are off unless you explicitly enable them, with a notice on each save.

Could my ID or government document numbers end up in memory?
Never; identification numbers, criminal history, and immigration status are prohibited from storage even with sensitive topics enabled.

Is the feature available in Saudi Arabia and the Gulf?
The announcement lists no regional restrictions; it is available globally for Free, Pro, and Max plans by default, while Team and Enterprise await admin activation.

For our complete guide to Claude and its updates, see the Claude AI guide, or compare options in Truescho's AI tools directory.

Sources

• Claude Blog — Claude's memory works everywhere, and you decide what's in it
• 9to5Mac — Anthropic update unifies memory feature across Claude Cowork and chat