1Password Lets AI Agents Log In Without Ever Seeing the Password

Password manager 1Password has launched "1Password for Agents." An AI agent gets a one-time digital key to sign into a third-party service, but it never sees the actual password. 1Password fills in the login itself, after the user approves the request on their own device. The process looks a lot like the two-factor checks people already use to add a layer of protection to logins. The feature already works with several AI agents: Meta's Muse, OpenAI's ChatGPT Work and Dots, Anthropic's Claude Duet, xAI's Grok Bot, and an agent called Town. 1Password CEO Aashish Tripathi put it bluntly: "the old model, where you hand over credentials and hope for the best, doesn't work."

This gives agents a working pattern for logging into other people's services. Access is scoped to one task and one session, instead of a standing password sitting somewhere. It only solves one problem, though. It settles who holds the login. It does not fix mistakes the agent can still make on its own, like booking the wrong thing or buying too much of something. And the list of supported agents already spans Anthropic, OpenAI, Meta, and xAI. This is not a single company's product.