Anthropic’s Claude 5 Sonnet update introduces persistent project memory that spans workspaces, so long-running initiatives keep context even when you switch between personal and team Projects.
How to turn it on
Memory is opt-in per Project via a new Memory tab. You choose what Claude may retain — style preferences, recurring stakeholders, glossary terms — and can audit or purge entries at any time. Nothing crosses Projects unless you explicitly link them.
The feature targets product teams and researchers who previously re-pasted the same brief every session. Early documentation emphasizes that memory complements, rather than replaces, uploaded files: docs still win for dense reference material.
Memory complements uploaded files — docs still win for dense reference material.
Anthropic also tightened default retention windows for free-tier workspaces and added export for memory entries, aligning with enterprise audit expectations.
