Two Claude updates landed this week, and together they change how the product feels: Anthropic merged the memory used by Claude chat and Claude Cowork into one shared system, and Cowork got a built-in browser that needs nothing to install. The first one fixes the most annoying thing about AI agents, the constant need to rebrief the assistant on things it already knows. The second one removes the last excuse for keeping Claude out of web tasks. Here is what both updates actually do, what Claude now remembers about you, and how to keep control of it.
What the Memory Unification Changes
Before this week, Claude chat and Claude Cowork kept separate memories. You could spend weeks developing an idea in chat, and the moment you handed the task to Cowork, you had to explain everything again: the headcount, the city, the speakers, the context. It felt like two products sharing a brand.
That is gone. Chat and Cowork now share one memory system, and it works both ways. What you discuss in chat is available to Cowork. What Cowork does is available back in chat. Anthropic describes it as one continuous assistant: the context you have built up across months of conversations is there the moment you hand Cowork a task, and anything that comes up in Cowork carries back to chat.
The mechanics changed too. Claude used to summarize a conversation after it ended and store the summary. Now it adds topics to memory as you chat. That means new memories are available immediately, mid-conversation, and they sync to Cowork without waiting for a session to close.
One important boundary: Claude Code is not part of this. When ZDNET asked, Anthropic confirmed the update is focused on Cowork and chat. Code keeps its own memory and context, so this does not change how Claude Code sessions behave.
What Claude Remembers (and What It Refuses To)
Memory is on by default across Free, Pro, and Max plans on web, desktop, and mobile. For the first time, Anthropic is also exposing what Claude has retained, so you can read, edit, or delete information on any topic. That alone is a big step forward: memory is no longer a black box you cannot inspect.
By default, Claude does not store sensitive topics: health data, race, ethnicity, religious beliefs, politics, gender identity, and similar categories. If you want Claude to remember that you have a gluten allergy so it can adjust recipe suggestions, you can toggle on “include sensitive topics in memory.” When a sensitive topic does get saved, the app tells you.
Some things are never saved, no matter what you toggle: government-issued IDs, Social Security numbers, criminal history, immigration status, and anything that violates Anthropic’s acceptable use policy. Those exclusions are not optional.
How to Control What Claude Keeps
The controls are straightforward and worth knowing before the memory builds up:
You can also pause memory entirely, which stops new topics from being saved while keeping what is already there, or reset it, which clears everything. There is no “off switch that forgets selectively,” so if you want a clean slate, reset and let Claude rebuild from your next conversations.
One practical tip: treat the memory manager like a browser history view. Every few weeks, open it and delete anything that is stale, wrong, or simply too personal to keep. A periodic cleanup keeps Claude useful without turning memory into a permanent dossier.
Cowork’s Built-In Browser: Nothing to Install
The second update is smaller but removes a real friction point. Claude Cowork now has a browser built into the desktop app. When a task needs a website, a browser opens in the side panel and Claude navigates pages, reads them, clicks, and types. No Chrome extension, no setup, nothing to install.
Previously, web tasks in Cowork meant giving Claude access to your browser through the Claude in Chrome extension. That is still the right choice when the work involves a page you already have open. But many web tasks do not need your browser, they just need a browser. Now Claude has one of its own.
The rollout details matter if you are on a specific plan: the built-in browser is rolling out this week to Pro, Max, and Team plans in the Claude desktop app. The browser lives in the desktop app, so Claude Desktop needs to be open and online for it to work, even though the Cowork session itself runs in the cloud. If you start a task on desktop, you can keep steering it from Claude on web or mobile while the desktop app stays open.
What This Means for Your Workflow
Taken together, these two updates close the loop between thinking and doing. The research phase and the action phase were previously split across two products with two memories. Now they are one.
A realistic example: you brainstorm a conference agenda in chat over a week. When you ask Cowork to build the budget and logistics doc, it already knows the headcount, the city, and the speakers. No rebriefing. You explain once in chat how your team defines its metrics, and every quarterly business review deck Cowork builds afterward uses those definitions.
For teams, the browser matters just as much. Cowork can now go to a competitor’s site, pull pricing, compare plans, and drop the results into a doc, all without you babysitting a browser session. Tasks that used to be “annoying but not worth automating” become worth automating.
The tradeoff is the one ZDNET flagged: a more useful assistant is also a more personal one. Memory that spans chat and Cowork is more powerful, and it is more sensitive. The controls above are the mitigation, and using them on a schedule is the habit.
OpenClaw agents keep persistent memory across sessions, and you control exactly what they retain. If your agent keeps forgetting context, our fix guide walks through the memory settings step by step.
Frequently Asked Questions
Yes. Anthropic merged the memory systems of Claude chat and Claude Cowork. What you discuss in chat is available to Cowork and vice versa, so you no longer have to re-explain context between the two.
Yes. The update exposes Claude’s memory so you can read, edit, or delete information on any topic. You can also pause memory or reset it completely.
Not by default. Claude does not store health data, race, ethnicity, religious beliefs, politics, or gender identity unless you toggle on the option to include sensitive topics. Some things are never saved, such as government-issued IDs and Social Security numbers.
No. The built-in browser lives in the Claude desktop app and needs nothing to install. Claude opens a browser in the side panel, navigates pages, reads, clicks, and types. The desktop app must be open and online for it to work.
No. Anthropic confirmed the memory unification is focused on Claude chat and Cowork. Claude Code keeps its own separate memory and context.
Claude just got a serious memory upgrade. See how the latest Claude models compare for real work before you commit.
Final Thoughts
Memory is the feature that turns a chatbot into an assistant. With this update, Claude finally behaves like one product with one context, and Cowork’s built-in browser means that context can go out and act on the web. The cost is that your assistant now knows more about you, so the memory manager is not a nice-to-have anymore. It is the control panel for how much of you lives in the machine. Check it, edit it, and reset it when the context gets stale. That is the whole game.
About The Author
