Claude's memory is now unified across Chat and Cowork

There are few things more annoying than spending months training an AI on your work habits, only to switch tabs and find it has complete amnesia about who you are.

On Tuesday, Anthropic solved this exact headache by merging Claude's chat memory with Claude Cowork. Context built up across your everyday conversations now flows directly into your autonomous workspaces, and vice versa.

Real-time memory you can actually edit

In practice, this means if you brainstorm a conference in chat, Cowork already knows the city, headcount, and speakers when tasked with building the budget doc. If you explain how your team defines metrics once, Cowork will use them across every quarterly business review deck moving forward.

Anthropic also changed how Claude takes notes. Instead of waiting for a session to end to write a summary, Claude now extracts topics in real time as you chat. If you mention that a deadline slipped to September, the next exchange already reflects that without you needing to say "remember this."

Even better, your memories aren't locked inside a black box. In Settings under Topics, you can view, edit, or delete every single note Claude holds. Fixing an outdated company name in one short file instantly updates its behavior across all future chats.

The guardrails on sensitive data

By default, Claude ignores personal and sensitive categories, including health data, religious beliefs, race, and politics. But if you want Claude to remember your gluten allergy for meal planning, you can toggle that permission on, and Claude will alert you every time it saves a sensitive detail.

However, some things are strictly banned regardless of your settings, including Social Security numbers, government IDs, criminal history, and immigration status.

The unified memory is already live and on by default for Free, Pro, and Max users across web, desktop, and mobile apps. For Team and Enterprise accounts, it remains off by default under administrative control.

Why it matters

We think this is one of the most practical UX upgrades Anthropic has shipped all year. It directly tackles the massive friction between the research phase and the action phase of working with AI.

Until now, most AI platforms felt like fragmented suites where the chatbot didn't talk to the workflow agent. Bridging that gap turns Claude from a series of disjointed tools into a single, continuous colleague that actually remembers context across your entire workday.

Giving users direct editorial control over individual memory files is also a brilliant move. Instead of praying an agent forgets a hallucinated detail, users can simply open the file and hit delete. The only real risk is real-time ingestion: if Claude picks up a passing hypothetical as an absolute truth mid-chat, you might find yourself doing a bit of memory hygiene down the road.