News
1Password for Agents lets AI agents log in to accounts without seeing your password
1Password has launched 1Password for Agents. An AI agent requests a login, the user approves it, and the password manager enters the credentials. Two agents work at launch, and access is through a waitlist.
1Password unveiled 1Password for Agents on October 22, 2026. The service is meant to let personal AI agents log in to a user's accounts without access to their password. The agent never receives the credentials: after a human approves the request, the password manager enters them on its behalf.
How it works
According to TechCrunch, an agent trying to log in sends 1Password a unique digital key, which the service uses to verify the agent's identity. The user gets an alert and can personally approve the attempt. Once they agree, 1Password fills in the credentials itself, so the agent never handles them.
9to5Mac quotes the company's announcement: 1Password will connect to a user's personal agents, one or several, so they can work without seeing the user's passwords. The company says this keeps credentials and secrets secure while the agents get tasks done.
Which agents, and when
TechCrunch lists Anthropic's Claude Duet, OpenAI's Dots and ChatGPT Work, Grok Bot, Meta's Muse, and agents from the startup Town among the supported agents. 1Password says it is working with other companies to expand the list.
9to5Mac describes the current state more cautiously: at launch only Muse and Grok Bot work, and integrations with OpenAI and Anthropic are coming later. Access is through a waitlist on the 1Password website. The company's announcement also names Browserbase, Browser Use, Perplexity, Ramp and other partners.
According to 9to5Mac, 1Password said the following about how the service connects to agents: "1Password will connect to your personal agents, whether you have one or multiple, so they can work without seeing your passwords."
Why it matters
In a blog post quoted by TechCrunch, 1Password says that handing passwords to agents is a common shortcut that can leave sensitive information in places users can no longer control. The company pitches its tool as a trusted way to delegate tasks to personal AI agents.
An Axios report cited by TechCrunch adds context: U.S. banks are detecting agent activity inside customer accounts, and most cannot tell whether the user requested it. A Chase spokesperson told Axios that customers should not give their usernames and passwords to anyone. TechCrunch notes this does not necessarily mean the agents are malicious, but the potential for problems is large.
Caveats
This is a vendor announcement. The description of the mechanism comes from the company, and neither source describes independent security testing of 1Password's approach. The sources also give no price or regional availability, so it is unclear when the service will reach users in Poland.
