What happened
1Password launched (July 16, 2026) a browser integration for Anthropic's Claude that lets the AI agent use stored credentials without the password, username, or MFA codes ever entering the model's context, memory, or Anthropic's systems. A new 'Agentic Mode' locks down the vault when an AI agent takes control, granting per-task, biometric-approved access only to explicitly authorized items.
Why it matters
This is the first major password manager to ship a dedicated secure channel for agentic credential use rather than pasting secrets into a model's context window, directly addressing a widely-cited gap (68% of enterprises can't distinguish agent activity from human activity) and setting a likely reference architecture for AI agent authentication.
Applicability
Enterprises deploying Claude-based browser agents (and other agent vendors as support expands) should evaluate zero-exposure credential injection now; IAM/security architecture teams should audit existing agent-credential flows against this pattern.