1Password for Agents review: your agent can sign in, and the password stays in the vault
Your assistant hits a login page and stops. Until this week the choices were to take over and type it yourself, or paste the password into a chat. 1Password launched a third option on October 22. It fixes the password problem. It doesn't decide what the agent does once it's inside.
1Password for Agents is worth turning on if you already keep your logins in a 1Password Individual or Family account and you use a supported personal agent. You approve each login on your own device, 1Password fills it in, and the agent never sees the password. It doesn't limit what the agent can do once it's signed in, Business accounts aren't covered yet, and most integrations are still marked coming soon. Teams should keep agents on their own lower-permission logins.
The verdict first
If you already keep your passwords in a 1Password Individual or Family account, and you've started handing errands to a personal AI agent, turn this on when your agent supports it. It's a real improvement on the two things people do today.
If you run a team, or your business logins live in a 1Password Business account, wait. The launch describes Individual and Family accounts only, and it doesn't touch the bigger question a business has to answer: what the agent is allowed to do once it's signed in as you.
That's the line to keep in mind for the rest of this review. Keeping the password away from the agent is not the same as keeping the account away from the agent.
What 1Password says it does
1Password launched the product on October 22, 2026. The problem it solves is simple to picture. Your agent is booking something, ordering supplies or filing an expense, and it reaches a sign-in page. Until now you had two options. You could take over the session and type the password yourself every time, which is slow. Or you could paste the password into the chat, which gives the agent a copy you can't take back.
Here's the new flow, as 1Password describes it. The agent stops at the login and asks for the credential. It proves its identity to 1Password with its own key. You get a request on your own device showing which login it wants and why. If you approve, 1Password fills the password or one-time code directly into the sign-in form. The agent only learns that access was granted. If more than one saved login matches, you choose which one it uses.
Engadget reported that access is granted per session, limited to the items you approved, and not carried over. Your other saved logins stay encrypted, and 1Password says connecting an agent doesn't open the rest of the vault. If you change a password, you change it once and connected agents use the new one. You can disconnect an agent at any time.
How the work moves
What happens at the login screen
- Agent stops
It reaches a sign-in page and asks for a credential.
- Agent proves who it is
It verifies itself to 1Password with its own key.
- You decide
Your device shows which login and why. Approve or deny.
- 1Password fills it
Password or one-time code goes into the form, not to the agent.
- Agent carries on
It now acts inside the account as you.
Keeping the password away from the agent is not the same as keeping the account away from the agent.
Where it falls short today
It stops at the front door. 1Password's own FAQ is direct about this: once signed in, "an agent may be able to see information or take actions within the signed-in account." If you approve a login to your supplier portal, the agent can do whatever you could do there. Engadget made the same point more bluntly, noting that agents can still make ordering mistakes.
Few agents work yet. At launch, Meta's Muse is listed as rolling out over the coming days. Integrations for Claude Duet, ChatGPT Work and Dots, and Town's Townie are marked coming soon, with Browserbase, Browser Use, Perplexity, Ramp and others after that. There's a waitlist for everything else. 1Password says safeguards and controls vary by integration, so the same approval may behave differently in two agents.
Approval frequency varies. How often you're asked to approve depends on the integration. That matters, because the protection depends on you actually reading each request instead of tapping yes.
It isn't a business product yet. Coverage is Individual and Family accounts on 1Password.com. Accounts on 1Password.eu and 1Password.ca aren't supported, and the launch materials don't describe Business accounts.
Who should wait
Banks are a good reason for caution. TechCrunch, citing Axios, reported on launch day that US banks have started to detect agent activity inside customer accounts, and that a Chase spokesperson said customers "should not give their usernames and passwords to anyone." 1Password's design means you aren't handing over the password, but the bank sees an agent acting in your account either way. Read your bank's terms before you approve a banking login for any agent.
Teams should wait too. The cleanest setup for a business is still a separate login for the agent with a smaller role, wherever the site allows one. An agent signed in as the owner of your Shopify store, your payroll or your ad account carries every permission the owner has. If a site lets you add a staff account with limited rights, create one for the agent, save that login, and approve only that one.
And if your agent work is mostly API keys and payment cards rather than website logins, look at the alternatives. Our Agent Vault review covers a different approach for secrets. Services that offer their own connector, with scopes you pick when you sign in, are usually tighter still, because the service decides what the connector can touch.
Decision guide
Approve it now, or set it up differently
| Consideration | Fine to approve | Set up differently first |
|---|---|---|
| Bookings | Restaurant, travel or event sites tied to you. | Company travel accounts shared by staff. |
| Supplies | A store account with no saved payment card. | Wholesale portals with credit terms on file. |
| Store admin | A staff login you created for the agent. | The owner login with every permission. |
| Banking | Nothing until your bank says agents are allowed. | Ask the bank, or keep doing it yourself. |
Price and plan
There's no extra charge. 1Password says the feature is available at no additional cost on Individual and Family accounts on 1Password.com, and it's included in those plans. You still need a supported agent, and each agent has its own plan and limits.
Since there's no separate price, the cost to think about is attention. Every approval is a small decision. Set yourself one rule before you start: approve a login only when you know what task it's for, and decline anything that surprises you.
A sensible first week looks like this. Connect one agent. Approve logins only for low-stakes accounts, like a restaurant booking site or a supplies store with no saved card. Watch what the agent does after it signs in. Add a more important account only once you're comfortable with how it behaves, and keep money-moving accounts out until your bank says it's fine.
Should you turn it on
Yes, if you're an owner with a personal 1Password account and a supported agent, and today you're either pasting passwords into chats or giving up on agents at every login screen. This is better than both.
Not yet, if your important logins are business accounts shared across a team, or if the agent would sign in as the account owner somewhere that touches money, customers or payroll. Give the agent its own limited login first, then use 1Password to keep that password out of the chat.
If you're building agent features into your own product and want to work out where approvals, separate logins and scoped connectors belong, that's the work our AI applications and agents team does. Talk to us.
Related work: Sutton Place.
We build alongside the people doing the work.
Bring us the problem you’re thinking about.