Claude's memory now follows users between chat and Cowork

Anthropic has merged the memory systems behind Claude's chat interface and its Cowork workspace tool, ending a long-standing annoyance for people who switch between the two. Announced August 25, 2026, the update means a project detail, preference, or piece of context mentioned in one interface now carries over automatically to the other.

Previously, Claude treated chat and Cowork as separate environments with no shared memory. Users had to repeat background information every time they moved between them, even mid-project. That re-briefing step is now gone.

The underlying mechanism has also changed. Rather than summarizing a conversation once it ends, Claude now adds relevant details to memory continuously as the chat unfolds. That shift lets the assistant reference earlier context sooner, without waiting for a session to close.

What Claude will and won't remember

Anthropic built in specific guardrails around sensitive data. By default, Claude excludes health information, race, ethnicity, religious beliefs, political views, and gender identity from memory.

A hard limit sits underneath that default: government-issued IDs, Social Security numbers, criminal history, and immigration status are never stored, regardless of settings. Content that breaches Anthropic's acceptable use policy is excluded as well.

Users aren't locked out of the sensitive categories entirely. A toggle lets people opt in to saving those topics if they choose, and Claude flags it when a sensitive detail gets written to memory.

Control over what's remembered extends further: users can read, edit, or delete anything stored in Claude's memory, on any topic, sensitive or not. That gives people a direct way to audit what the assistant has retained about them.

The feature is switched on by default across Free, Pro, and Max subscription tiers, spanning web, desktop, and mobile. Mobile users on iOS and Android will need to update to the latest app version before the merged memory becomes available to them.

For teams already running Cowork alongside regular Claude chat, the practical effect is fewer repeated instructions and less friction when moving a task from a quick chat exchange into a structured Cowork session. Anthropic's framing suggests this is aimed at making Claude feel like a single continuous assistant rather than two separate tools sharing a name.