Claude Unifies Memory Across Chat and Cowork

Users no longer face the friction of restarting their workflow. Previously, information shared during general brainstorming sessions remained siloed, forcing a manual transfer of data when moving to the Cowork agent. Now, Claude functions as a single, continuous assistant. For example, if you discuss conference logistics—such as headcount or speaker lists—in chat, Cowork will automatically access that data to draft agendas without further prompting.

Anthropic has shifted to a real-time memory model, updating retained topics as conversations happen rather than waiting for a session to conclude. To maintain transparency, the company is exposing these memory banks, granting users the ability to read, edit, or delete stored information at will. While the system is enabled by default across Free, Pro, and Max tiers, it includes strict guardrails: sensitive categories like health data or political beliefs are excluded unless the user explicitly toggles them on. Regardless of settings, Claude remains prohibited from storing high-risk data such as Social Security numbers or government-issued IDs.