1Password is rolling out a new integration that lets Anthropic’s Claude AI agent sign into websites using credentials stored in 1Password, without Claude seeing those secrets. Multiple outlets report that the feature is aimed at Mac users and works through a browser integration. With user authorization, Claude can access stored information such as usernames and passwords to complete multi-step online tasks, including actions like booking travel and managing account-related workflows. Sources also say the integration is designed to handle authentication flows while keeping sensitive data protected: the system enables Claude to enter passwords and two-factor authentication (MFA) codes as needed, but does so in a way that prevents the model or Anthropic from directly receiving the credential values. ZDNet and other coverage describe this as part of an “agentic” approach, positioning 1Password as a trusted identity and access layer between the user’s stored credentials and the AI agent performing the tasks. The announcements emphasize that users remain in control by approving access to their accounts before Claude can use any stored security details.