1Password has rolled out a new integration that lets Anthropic’s Claude AI access stored usernames, passwords, and other credentials directly from user vaults, all without the sensitive data ever touching Anthropic’s models. The feature, part of 1Password’s broader Unified Access platform launched on March 17, is designed to let AI agents handle multi-step tasks like booking travel or managing online accounts on a user’s behalf.

How zero-exposure actually works

The integration relies on what 1Password calls a “zero-exposure security framework.” Instead of sending your credentials to Claude’s servers, 1Password injects them directly at runtime. The AI agent can use your login to complete a task, but it never actually sees or stores the password itself.

The integration works across multiple surfaces, including the Claude browser extension, Cowork, and Claude Code. Each interaction requires explicit user consent before any credential is accessed.

1Password currently protects over 1.3 billion credentials across more than 180,000 businesses and serves over 1 million developers.

The bigger picture: AI agents need identity rails

1Password’s ecosystem partners at launch include not just Anthropic but also OpenAI and GitHub. The platform also focuses on securing API keys and Model Context Protocol (MCP) environments, which are the plumbing that connects AI tools to external services.

1Password’s broader Unified Access platform emphasizes three pillars: discovery, governance, and audit. Discovery means identifying all credentials in use across an organization. Governance means controlling who, or what, can access them. Audit means logging every interaction for compliance.