1Password for Agents Logs an AI In Without Your Password. It Doesn’t Stop It Once It’s Inside.
TL;DR: On 22 October 1Password launched 1Password for Agents: when an AI agent hits a login page it stops, asks 1Password for the credential, and you approve — then 1Password types the password into the site itself, so the agent never holds it. It is free on Individual and Family accounts. Two things to know before you rely on it: Claude Duet is not live yet, and 1password for agents stops an agent holding your password, not an agent acting once the door is open.
What shipped
1Password’s announcement of 1Password for Agents describes a three-step flow. The agent reaches a sign-in page and stops. It asks 1Password for the credential it needs. You review the request and approve it, choosing which stored login it may use — and then 1Password “directly auto-fills the credentials in the destination site so the agent doesn’t have access to the credentials.”
TechCrunch adds the identity half: the agent proves which agent it is with its own digital key before the request reaches you, so the prompt tells you who is asking as well as what for.
1Password’s chief product officer Aashish Tripathi gave TNW the clearest statement of intent: “An agent should be able to do the job it’s been asked to do without getting permanent access to everything behind the account.”
Four things 1Password for Agents does and doesn’t do
1. It stops the agent holding your password. It does not stop the agent using your account.
This is the distinction almost every write-up of 1Password for Agents skipped. Once 1Password has filled the form, the agent is signed in and inside, with whatever that account can do. The protection is real and it is narrow: it removes a copy of your credential from a system you don’t control. It does not add a second gate around what happens next. Keep the two questions separate — “who holds my password” and “what can this thing do once it’s in” are different, and only the first one was answered this week.
2. Claude Duet isn’t live on it yet.
If you read yesterday’s piece on Claude Duet and went looking for 1Password for Agents inside it, stop looking. 1Password’s own page puts Muse and SpaceXSI’s Grok Bot in “rolling out now”, “over the coming days across all platforms”, and lists Claude Duet, OpenAI’s ChatGPT Work and dots, and Town’s Townie as coming soon, with no dates. Perplexity, Ramp, Browserbase, Browser Use and several others are queued behind them.
3. “Safeguards and controls vary by integration.”
That is 1Password’s own wording, along with “availability, platforms, and capabilities vary by partner integration.” It is admirably honest and it is also the thing to watch: the single name 1Password for Agents is about to cover a dozen different security postures. We saw the same shape when Meta’s Muse share links leaked — one brand, two products, two different answers to “who has my data”. Check what your agent’s integration actually does rather than trusting the badge.
4. Nobody has said how long “yes” lasts.
The 1Password for Agents page describes approving each requested credential and says you can “disconnect the integration at any time”. It does not say whether an approval covers one sign-in, one task, one session, or stands until you revoke it. For something that logs into your accounts while you are elsewhere, that is the question — and it is unanswered.
Why this matters if you don’t code
Here is the week in one line. On Monday we wrote that an API key is a payment instrument with no printed limit. Yesterday, Claude Duet arrived as an agent that works in your accounts while your laptop is shut. The obvious question — so do I give it my passwords? — now has a proper answer instead of a shrug, and 1Password for Agents is it.
The old answer was to paste the password in and hope. 1Password’s framing is blunt about why that was bad: handing over credentials is “a common shortcut” that leaves sensitive information where you can no longer control it. TechCrunch, citing Axios, reports that US banks have begun detecting agent activity inside customer accounts and mostly cannot tell whether it was something the customer asked for.
So the practical rule for a non-developer is simpler than it looks. If an agent asks you to type a password into it, that is now a choice you are making rather than the only option available — and on the accounts that matter, it is the wrong one.
Who should care (and who shouldn’t)
How much of 1Password for Agents applies to you depends on what you are running.
- Running a small business: care most, and check your region first. 1Password says accounts on 1Password.eu or 1Password.ca aren’t currently supported, which will catch a lot of UK and European users.
- Building a revenue app: useful for the admin accounts around the product. The agent still acts with full rights inside whatever it signs into, so start where a mistake is cheap.
- An internal tool for your business: this launch is for Individual and Family accounts. Team and business plans are not what was announced.
- Learning: nothing to do. You do not need agents signing into real accounts to learn how any of this works.
- Already using Muse: you are the one group who can try this today.
Our take
1Password for Agents is the right shape, and we would turn it on. Removing the password from the agent’s memory is a genuine structural fix rather than a policy promise, and 1Password did the unglamorous thing of saying plainly that protection varies by partner instead of letting the brand imply uniformity.
What we would not do is file it under “sorted”. 1password for agents solves credential custody. It leaves the harder problem exactly where it was: an approval prompt, shown to a busy human, is still the only thing between an agent and an account — and we have written before about how reliably people wave those through. A better lock on the door is worth having. It is not the same as knowing what is happening in the room.
Not sure which AI tools are worth connecting to anything of yours? Take the 60-second Vibe Coding Tool Finder quiz →
Some links on this page are affiliate links. If you buy through them we may earn a commission at no extra cost to you.
FAQ
Does 1Password for Agents cost extra?
No. 1Password says it is available “at no additional cost for 1Password Individual and Family accounts on 1Password.com”. Accounts on 1Password.eu or 1Password.ca are not currently supported, and the announcement covers personal plans rather than business ones.
Which AI agents work with 1Password for Agents today?
1Password lists Meta’s Muse and SpaceXSI’s Grok Bot as rolling out now, over the coming days. Claude Duet, OpenAI’s ChatGPT Work and dots, and Town’s Townie are listed as coming soon with no dates given.
Can the agent still see my password?
1Password says no — it fills the credential into the website itself so the password is never exposed to the agent or the model. It can still do whatever the account allows once signed in, which is a separate question.
