1Password has launched "1Password for Claude", a browser extension feature that lets Anthropic's AI agent use stored login credentials to carry out tasks such as booking travel or managing accounts, without those credentials ever reaching Claude, its memory or Anthropic's servers. The launch addresses a growing security concern around AI agents: sharing passwords directly with a model exposes them to potential compromise if the model or its underlying systems are hacked, so 1Password's approach aims to let users delegate real-world tasks to Claude without giving up control of sensitive credentials.

The feature relies on a "zero-exposure security framework" that injects passwords and one-time MFA codes through a secure 1Password channel, with access granted per session and limited to specific, pre-approved items rather than carried over indefinitely. It is accompanied by a new "Agentic Mode" for all 1Password users, which locks down credential access whenever an AI agent is detected and scans pages after each autofill to confirm nothing remains exposed. The Claude integration is now available to 1Password business, family and individual plan users on Mac, requiring both the 1Password and Claude desktop apps and browser extensions, with support for payment cards and identity details, plus other AI agents beyond Claude, promised for the future.