On a Thursday, 1Password introduced 1Password for Agents, a service that lets people authorize AI agents to sign into their personal accounts without handing over the actual password. The announcement landed the same day as an Axios report describing how U.S. banks are detecting automated agent activity inside customer accounts.
How a login works without a password
The design keeps credentials inside 1Password throughout the process. When an agent needs access, it sends a unique digital key to 1Password so the password manager can verify which agent is making the request. The user then receives an alert about the attempt and can personally confirm it.
Only after that confirmation does 1Password sign the agent in and fill the credentials on its behalf. The agent gets a working session and completes whatever task it was assigned, but never receives the underlying username or password. The system sits between the agent, the user and the password manager, mediating the exchange rather than passing secrets along.
Which agents are supported at launch
The service supports agents from several companies at launch. Users can choose among Claude Duet, Dots, ChatGPT Work, Grok Bot, Muse and Town agents. Those names span Anthropic, OpenAI, Meta and Town, giving the tool a fairly broad starting roster.
1Password says it is working with other companies to expand the list. The company has not disclosed pricing or general availability details beyond the launch announcement.
Why 1Password built it
In a blog post, the company framed the tool as a response to a common but risky habit: people hand their credentials directly to automated services so those services can act for them. Once copied, 1Password warns, credentials can end up in places users no longer control.
The company positions the new service as a trusted way to delegate tasks while retaining control over credentials. That rationale matters as more consumers let AI agents handle account access, a shift that security services are now trying to accommodate.
Banks see agentic activity they cannot fully explain
The launch coincided with an Axios report on U.S. banks detecting agentic activity in customer accounts. Much of that activity reflects consumers allowing automated services to act on their behalf, according to the report. But banks reportedly cannot tell whether a given agentic action is tied to a user’s request or not.
Bankers described the situation as horrifying, Axios reported. The concern is not that every agent action is malicious, but that financial institutions lack a reliable way to distinguish authorized automation from unauthorized use.
Chase's position on sharing credentials
A Chase spokesperson told Axios that customers should not share usernames and passwords. The warning is straightforward and aligns with the broader problem 1Password is trying to address: the more agents act inside accounts, the more valuable it becomes to keep passwords out of their hands.
1Password for Agents does not resolve the banks’ detection problem on its own. What it offers is a structure in which the agent’s identity is verified, the user approves each login, and the credential never leaves the password manager. Whether that model becomes standard for agent access is still an open question, but the launch gives consumers a concrete option beyond copying passwords into an AI tool.
[…] 1Password Lets AI Agents Log In Without Ever Seeing Your Password […]