Claude can now send emails in Gmail, even without your approval

Anthropic's Claude can now send Gmail messages autonomously, without prompting the user for approval before each action.

This is a meaningful threshold crossed: the model moves from suggesting actions to executing them inside a live communication channel. The difference between drafting and sending is the difference between a tool and an agent with real-world consequences.

The full piece is worth reading for the specifics on how permissions are scoped and what guardrails, if any, Anthropic has built around unsupervised send behavior. That implementation detail is the story.