Updated
Updated · ZDNet · Jul 16
1Password Launches Agentic Mode for Claude, Hiding Passwords and MFA Codes
Updated
Updated · ZDNet · Jul 16

1Password Launches Agentic Mode for Claude, Hiding Passwords and MFA Codes

3 articles · Updated · ZDNet · Jul 16

Summary

  • Agentic Mode is now available to all 1Password users, letting Claude access protected accounts by having 1Password inject passwords and one-time MFA codes directly into web pages.
  • Claude never sees the credentials or receives them in its context, 1Password said; each access request is tied to a specific task and requires user approval through biometrics, Touch ID or a password.
  • 1Password said the permission does not create standing access: if Claude needs to revisit a site after logout, it must request approval again, though normal browser sessions may persist until the user signs out.
  • The feature launches first with Claude and runs automatically when a recognized AI agent takes over the browser, with support for payment cards, identity details and more browser-based agents planned later.

Insights

Could making AI access so 'secure' lead users to carelessly approve requests, creating an entirely new security vulnerability?
Is this human-in-the-loop model just a temporary fix until AIs can be trusted with their own digital identities?
If an AI agent commits a costly error after you approve its access, who is legally and financially responsible?