1Password lets Claude use your credentials without seeing them

1Password has introduced a browser integration for Claude that enables Anthropic's AI agent to use stored credentials (usernames, passwords, MFA codes) without ever seeing them. The feature relies on a zero-exposure security framework: when Claude requests access for a specific task, 1Password injects the required credentials through a secure channel that the AI cannot read. Each request requires user approval via a single biometric prompt, minimizing workflow interruptions while maintaining control. After autofill, 1Password scans the page to ensure no credentials remain exposed before returning control to Claude. The vault stays locked except for the explicitly granted credentials, preventing broader access.

The integration is available now for 1Password users on Mac (desktop app and browser extension required, alongside Claude’s desktop app and extension). It supports business, family, and individual plans. Initially, Claude can only access login-related credentials (passwords, passkeys, 2FA codes, API tokens). Support for payment cards and identity details is planned for a later update. This move addresses a key pain point in AI agent usage: the need to manually enter credentials for each step of a multi-step task. By enabling secure, per-task credential injection, 1Password reduces friction while keeping sensitive data hidden from the AI model itself.