1Password Enables Claude to Access Logins Securely

A new integration between Anthropic's Claude AI and the 1Password password manager allows users to automate online tasks that require logins without exposing their credentials to the artificial intelligence. The feature, which launched on July 16, 2026, is currently available for Mac users.

The integration enables Claude to perform actions like online shopping, booking travel, or managing accounts by securely accessing stored usernames and passwords. The core security mechanism ensures Claude never directly sees the login information. When the AI needs to sign into a website, it requests the credentials from the 1Password connector. 1Password then injects the username, password, and any required one-time authentication codes directly into the target website through a separate, secure channel. This data is kept out of Claude's context, memory, and Anthropic's systems.

Access is granted on a per-session basis, with users approving or denying each credential request via a biometric prompt. The permission lasts only for the current task and does not leave any standing access. To further enhance security, 1Password has introduced a feature called Agentic Mode. When an AI agent like...