The password problem is becoming an agent problem. 1Password has launched a way for AI agents to log in on a user’s behalf without seeing the user’s password, TechCrunch reported on Thursday.

The design goal is the one that has defined password managers from the start, now applied to software actors: the credential should be used without being exposed. An agent that can complete a sign-in without ever holding the secret in plain view has less to leak, less to log by accident, and less for an attacker to steal if the agent itself is compromised.

The launch lands in a week when agents stopped being a demo category. Anthropic’s Claude Duet began working in the background for professionals, OpenAI’s tools spread through large employers, and TechCrunch’s own daily list was dominated by agent products — each of which will, sooner or later, need to get through a login screen.

That is why the identity layer matters. Browsers and operating systems solved sign-in for humans over two decades; the industry is now racing to build the equivalent for agents, with approval, scoping and secrecy built in rather than bolted on.