This is the direct answer to a problem I’ve been harping on all week. 1Password announced 1Password for Agents, a platform that lets AI agents use your logins without ever seeing or storing the passwords, MacRumors reported.
How it works
- An agent sends a request to sign in to a website or account, and you approve or deny it.
- The agent proves who it is with a unique digital key, and the login details are filled directly into the sign-in form. The agent never gets the password itself.
- It’s rolling out “over the coming days” for Grok Bot and Muse (Meta’s agent). Claude Duet, ChatGPT Work and Dots, and Town’s Townie integrations are “coming soon.”
- The feature is free for Individual and Family accounts on 1Password.com. 1Password plans start at $3.99 per month with annual billing.
Why I care
When the trojaned LiteLLM release stole API keys and passwords, the lesson was simple: anything an agent can read, an attacker running inside that agent can read too. The better design is for the agent to never hold the secret in the first place. This does that for website logins, and the per-request approval gives you a chance to say no.
My take
Great direction, but it isn’t magic. If you approve every prompt without looking, you’ve just moved the weak spot to your finger. Read what the agent is asking to sign in to, and why, before you tap approve. And keep unique passwords and two-factor on the accounts you hand to agents.
Have a blessed day! And Happy Hacking!
Update (Oct 24): Omarchy is already on board. DHH added 1Password to Omarchy’s Install, Service menu, pointing to the new agent credential flow as a fit for an agent-native OS. (Aligned News)
