Tools AI Minute Newsroom 2026-08-26

Claude's memory now follows you from the chat window into the work app — and it is built to forget your politics unless you say otherwise

Anthropic said on 25 August that Claude's chat product and Cowork, its agentic work app, now share a single memory. Whatever Claude picked up in conversation is available when Cowork does the work, so a project talked through over several sessions no longer has to be re-explained before anything gets built. Memories are now formed during a conversation rather than reconstructed from it afterwards. Claude Code keeps a separate memory for the time being. By default the system does not store sensitive categories — health, ethnicity, religion, politics, gender identity — unless the user turns a toggle on. Users can browse what Claude remembers topic by topic, edit or delete individual entries, pause memory entirely, or use incognito chats that leave nothing behind. It is on by default for Free, Pro and Max, with Team and Enterprise administrators deciding for their organisations; Cowork itself is not available on the free tier. One consequence of the merge, which Anthropic states plainly: there is no longer a way to keep chat memories and Cowork memories apart, because they are now one store.

Why it mattersMemory is what turns a chatbot into something that accumulates. It is also the feature that quietly converts a tool you use into a file kept about you. The design choices here are the part worth noting — sensitive categories excluded by default rather than opt-out, and a memory you can read entry by entry — because they set a bar competitors will be measured against. The trade-off is the one the company admits: unify the memory and you lose the ability to keep the two contexts apart. If you were treating the work app and the chat window as separate rooms, they are now one room.

#AI Agents

✓ Verified · 4 sources

Read in the app — free, in 9 languages

The nameless model that burned through 42 trillion tokens in six days has an owner: Z.ai says Ox Alpha is a GLM, and the weights are opening
2026-08-26 An agent that keeps thinking after you stop typing — and the whole harness is under 10,000 lines of Bash
2026-08-26 Nvidia says its new rack does 30 times more agent work per megawatt — it ran the test itself, and the benchmark's authors have not signed off
2026-08-25 The new Mac Studio holds 512GB of memory the GPU can address directly — a rack's worth of model, under a desk
2026-08-25 ByteDance has folded its coding tool and its agent builder into one app, and pointed it at the office
2026-08-25

Claude's memory now follows you from the chat window into the work app — and it is built to forget your politics unless you say otherwise

Related stories