1Password lets AI agents sign in for you without ever seeing your password

Anyone who has tried to let an AI agent handle something on a website has hit the same wall: the sign-in page. Either you hand the AI your password, which is a terrible idea, or you log in yourself and the agent loses half its point. On October 22, 1Password announced a third option.

Here’s how it works. The agent reaches a login page and asks for access. You get the request in 1Password, check which login it wants, and approve it. 1Password then fills in the username, password and one-time code on the site itself. The agent carries on, signed in, without having seen any of it and without access to the rest of your vault. You can see what was shared and disconnect the integration at any time.

Which agents work today

At launch, three agents were live: Meta’s Muse, SpaceXSI’s Grok Bot and Town’s Townie. Anthropic’s Claude Duet and OpenAI’s ChatGPT Work and dots are listed as coming soon, along with Perplexity, Browserbase, Browser Use, Ramp and others.

There’s no extra charge for Individual and Family accounts on 1Password.com. Accounts hosted on 1Password.ca and 1Password.eu aren’t supported yet, so readers in Canada and the EU will have to wait.

What to watch

1Password itself notes that safeguards vary by integration. In plain terms, your security also depends on how each agent company built its side of the handoff. It’s worth reading the permissions carefully the first time you connect one.

I think this is the right approach, and maybe the only sane one. Agents that browse for you are going to be routine, and the alternative, pasting passwords into an AI chat, already happens more than it should. Approving every login adds a little friction, but it’s exactly the friction that protects you. The real test comes when Claude and ChatGPT join the list, because that’s where most users are.