Claude Cowork finally remembers what you told the app in chat

A alteration to Claude’s representation strategy volition destruct 1 of the astir annoying things astir utilizing agents — the changeless request to rebrief the AI connected things it already knows. On Tuesday, Anthropic announced it’s merging the representation strategy utilized by chat and Claude Cowork, which means that Claude volition ever retrieve what it learned successful 1 area, adjacent erstwhile you’re engaging with it successful another.

Anthropic is besides exposing what Claude has retained successful its memory, allowing users to read, edit, oregon delete accusation connected immoderate topic.

Image Credits:Anthropic

The update volition amended the acquisition for Claude’s users arsenic it volition let them to much seamlessly determination betwixt the probe signifier and taking action.

Before, you whitethorn person developed ideas for projects implicit clip portion chatting with Claude, but erstwhile it came clip to really enact the changes, you’d person to commencement implicit by telling Cowork each the details. This was incredibly frustrating. The update makes Claude present consciousness similar 1 continuous assistant, alternatively than 2 abstracted products nether 1 roof.

Anthropic offers an illustration of idiosyncratic needing to draught an update for their manager oregon an docket for a conference, wherever Cowork would already recognize the references due to the fact that of the past conversations astir the substance astatine hand. For instance, it would cognize the headcount, city, and speakers astatine the conference, truthful you wouldn’t person to re-input that information.

Image Credits:Anthropic

In addition, Claude volition present adhd topics to its representation arsenic you chat, alternatively of summarizing a speech erstwhile it ends. That makes it easier to determination backmost and distant betwixt chat and Cowork, arsenic the caller memories are much rapidly updated and shared betwixt the 2 experiences, adjacent if the chat is ongoing.

Anthropic says that, by default, Claude won’t store idiosyncratic oregon delicate accusation similar your wellness data, race, ethnicity, spiritual beliefs, politics, sex identity, and more. However, users tin opt to alteration these things by toggling connected “include delicate topics successful memory” if they prefer. The app volition besides notify users whenever a delicate taxable is saved successful memory.

However, the institution says that Claude won’t ever prevention definite things, similar government-issued ID and Social Security numbers, transgression history, migration status, oregon different things that interruption its acceptable usage policy.

The representation diagnostic is enabled by default connected Free, Pro, and Max plans crossed web, desktop, and mobile. On iOS and Android, users volition request to update their app to the latest version.