1Password for Agents: AI Logins Without the Password — and the Catch After Sign-In

Quick answer

1Password for Agents, launched October 22, 2026, lets supported AI agents sign in to your accounts without ever seeing your password. You approve each login on your own device, and 1Password fills the credential directly into the sign-in form. The feature costs nothing extra on Individual and Family plans on 1Password.com, but it does not limit what an agent can do once it is inside an account.

An agent that can plan a trip, reorder groceries or file an expense report is not much use if it stops at every login screen and waits for you to type a password. Until now, the workaround was blunt: take over and sign in yourself, which defeats the point of delegating, or paste a password into a chat window, where a copy sits in a history you no longer fully control.

On October 22, 2026, 1Password announced a third option. 1Password for Agents lets supported personal AI agents sign in using credentials stored in your 1Password vault, without the agent ever seeing or storing the password itself. The company calls it a trust layer for personal AI agents. That framing is worth testing, because the product solves one specific problem well and leaves another one untouched.

How a sign-in works, step by step

The flow, as described in 1Password's announcement and launch coverage, has five stages:

1. The agent identifies itself. When a supported agent reaches a sign-in form, it presents a unique digital key to 1Password, proving the request came from that agent rather than an arbitrary script. 1Password has not published the key's format or how it is issued, rotated or revoked.
2. The agent states what it needs. The request names the specific credential and the reason. You see which agent is asking, for which login, and why.
3. You approve or decline on your own device. Nothing proceeds until you respond. John Torrey, 1Password's chief business officer, told PCWorld that the agent cannot approve its own access, and that users decide which credentials an agent can use, for which task and for how long.
4. 1Password fills the form itself. It enters the username, password and any one-time code directly into the sign-in form through a channel it controls. Engadget reported that access is granted per session and applies only to the approved items.
5. The agent continues. From the agent's side, it only learns that access was granted and that it is now past the login screen.

Two limits are built in. Connecting an agent does not open your whole vault: agents can only use logins you have approved. And passwords saved in a browser's own password manager, or typed into notes, are not usable through this system, according to 1Password's launch blog. A login needs to be a proper 1Password vault item first. You can also disconnect an integration at any time.

Which agents work with it, and which do not yet

The launch roster names six agent products, but they are not all live on day one:

• Rolling out over the coming days: Meta's Muse and SpaceXSI's Grok Bot.
• Coming soon, with no date given: Anthropic's Claude Duet, OpenAI's ChatGPT Work and dots, and Town's Townie.
• Building integrations for later: Browserbase, Browser Use, Dana, Instinct, KERNEL, Perplexity, Ramp, Tomo and Wajo.

Two conspicuous absences were noted by PCWorld: Google's Gemini Spark and Microsoft's Scout are not on the launch roster, and 1Password has not explained why. There is also a waitlist for unsupported agents. The practical takeaway is simple: this is not a universal "any chatbot can sign in" feature. It works only with supported integrations, and for most headline names, it works soon rather than today.

1Password for Agents builds on 1Password for Claude, launched July 16, 2026, which worked only on the Mac, inside Chrome, alongside the desktop app and browser extension. The new platform aims at agents not tied to one desktop browser session, routed through a common identity and approval process.

What it costs, with the fine print

There is no separate price for 1Password for Agents. 1Password says the feature is included at no additional cost in Individual and Family plans held on 1Password.com. That sentence carries three qualifications worth reading carefully.

First, the subscription itself is not free. On 1Password's personal pricing page, as read on October 24, 2026, Individual costs $3.99 per month billed yearly ($47.88 a year) at the regular price, or $2.99 per month ($35.88 a year) under a first-year promotion for new customers subscribing directly through 1Password.com. Families, covering up to five people, costs $5.99 per month billed yearly ($71.88 a year), or $4.49 per month ($53.88 a year) in the first year on the same terms. The promotion is limited to the first year with annual billing and is not available through Google Play, the Apple App Store or other third-party app stores. There is no free plan; new users get a 14-day free trial.

Second, the account has to live on 1Password.com. Accounts hosted on 1Password.eu and 1Password.ca are not supported at launch, which matters for European and Canadian users whose accounts sit on those domains. 1Password has given no timeline for adding them.

Third, availability varies by integration, region, device and sign-in method, in 1Password's own words. Even with a qualifying plan, a given agent, device or site may not work yet.

The catch: hiding the password is not the same as limiting the agent

This is the part of the announcement that deserves the most attention, because it is easy to read past it. Signing in usually produces a session cookie or token that keeps you logged in. Once 1Password has filled the form, the agent is operating inside a logged-in session, and it can do whatever that session allows: place an order, change a setting, send a message. The password stays secret, but the account is still in the agent's hands for the duration of the task.

That is why 1Password's own framing stresses approval per task rather than claiming the agent is harmless once inside. The sensible rule is to approve narrow accounts, not powerful ones. A grocery order you can review before delivery is a different risk from a bank account or an email account that can reset every other password you own. PCWorld's reviewer, who had tested the earlier Claude integration, drew exactly that line: groceries he could review, yes; bank or medical portals, no. The same session risk runs through other agent stories we have covered, from Walmart's shopping agents with built-in spending limits to Mastercard's agent payment rails: the credential is only the first gate, and what the agent may spend or do afterwards is a separate control.

There is a second, related risk no password manager can fix on its own. Prompt injection, where hostile text on a web page steers an agent's behaviour, remains unsolved across every major agent product. Hiding the password removes one thing a hijacked agent can leak, but not its ability to act inside a session it legitimately holds. TechCrunch's launch coverage also noted, citing an Axios report, that US banks are already seeing agent activity inside customer accounts that they struggle to attribute to the customer.

Setting it up: a practical checklist

1. Start with a 1Password.com account. Individual or Family; .eu and .ca accounts are not supported yet.
2. Move delegated logins into your vault. Browser-saved passwords and notes will not work; each login needs to be a proper vault item with the correct address.
3. Connect only a supported agent. As of October 24, 2026, Muse or Grok Bot in the current rollout; the rest are coming soon.
4. Approve per task, not forever. Check the agent name, site and item on every prompt; deny anything you did not ask for.
5. Keep email and banking out for now. Start with low-stakes accounts and check afterwards for orders, sent messages and connected apps.
6. Revoke what you no longer use. Disconnect an integration when a task is done.

How it compares

Pros and cons

Pros: passwords and one-time codes never reach the agent or its provider; approval happens on your own device, per request; access is limited to approved logins, not the whole vault; no extra charge on Individual and Family plans; integrations can be disconnected.

Cons: most headline agents are still "coming soon"; .eu and .ca accounts are excluded with no timeline; availability varies by integration, region, device and sign-in method; session-level risk remains; key issuance, session duration, audit logs and passkey support are unpublished; no independent security review exists yet.

What is still unknown

Several important details are missing from the launch materials, and they are worth stating plainly rather than filling in. 1Password has not published how long an approved session lasts, whether there are audit logs showing every sign-in an agent performed, or how the agent identity keys are issued, stored and revoked. It has not said whether passkeys, which cannot simply be typed into a form, can be used by agents. Anthropic's Ami Vora said only that 1Password will "soon" complete sign-ins using existing passwords for Claude Duet, with no date. Nor is it clear how approval prompts will behave when an agent runs a long task touching many accounts, or what happens on services requiring a fresh verification step mid-session. Independent security reviews of the design have not yet been published.

Scale figures are company-reported, not independently audited: 1Password says its vaults protect more than 1.6 billion credentials and that over 200,000 businesses use it, describing the password manager rather than the new feature. Its Developer Pulse survey finding (65% of developers expected or encouraged to use AI agents, only 33% with a secure way to do so) is likewise the company's own survey.

FAQ

Does 1Password for Agents cost extra?

No. 1Password says it is included at no additional cost in Individual and Family plans on 1Password.com. You still need a subscription: Individual is $3.99 per month billed yearly at the regular price ($2.99 in the first year for new direct customers), and Families is $5.99 per month billed yearly ($4.49 in the first year), covering up to five people. There is a 14-day free trial and no free plan.

Which AI agents work with it today?

Meta's Muse and SpaceXSI's Grok Bot are rolling out over the coming days from the October 22 launch. Anthropic's Claude Duet, OpenAI's ChatGPT Work and dots, and Town's Townie are listed as coming soon. Other companies, including Browserbase, Browser Use, Perplexity and Ramp, are building integrations for later. It does not work with every chatbot.

Can the AI agent see my password or one-time codes?

According to 1Password, no. After you approve a request, 1Password enters the username, password and any one-time code into the website's sign-in form itself, through a channel it manages. The agent only learns that access was granted. Once logged in, however, the agent can act inside that session, so only approve accounts you are comfortable delegating.

Can I use it with a 1Password.eu or 1Password.ca account?

Not at launch. 1Password says accounts hosted on 1Password.eu and 1Password.ca are not supported, and it has given no timeline for adding them. The feature requires an account on 1Password.com.

How do I stop an agent from using my logins?

Decline any individual sign-in request when it appears, and remember that agents can only use logins you have specifically approved rather than your whole vault. 1Password also says you can disconnect an agent integration at any time. Because each approval is tied to a task rather than granted permanently, an agent should need to ask again for later work.

Key takeaways

• 1Password for Agents, launched October 22, 2026, lets supported AI agents sign in without ever seeing your password, with you approving each login on your own device.
• Muse and Grok Bot lead the rollout; Claude Duet, ChatGPT Work, dots and Townie are still listed as coming soon, and Gemini Spark and Microsoft Scout are absent.
• The feature costs nothing extra, but the subscription does: Individual is $3.99 a month billed yearly ($2.99 in year one for new direct customers), Families is $5.99 ($4.49 in year one), and 1Password.eu and 1Password.ca accounts are excluded.
• Hiding the password does not limit the session: once signed in, an agent can do whatever that account allows, so approve narrow, low-stakes accounts first.
• Key details remain unpublished, including session length, audit logs, identity-key handling and passkey support, and no independent security review exists yet.

1Password for Agents is a narrow, well-aimed fix: it replaces the worst delegation habit, pasting a password into a chat, with an approval step you control. It does not make an agent trustworthy inside your accounts. That judgement is still yours, one approval at a time, and as of October 24, 2026, most agents on the launch list have not even arrived yet.