The key facts

  • 1Password for Agents: the agent requests a login, the user approves it, and the password manager itself fills in the credentials
  • Agents from Anthropic, OpenAI, xAI, Meta and the startup Town are supported; the list is expected to expand
  • According to Axios, US banks are detecting agent activity in accounts and cannot always link it to user requests

A new way to delegate logins to agents

Password manager 1Password on October 22, 2026 introduced the 1Password for Agents platform. It allows users to let AI agents log into their personal accounts on their behalf, while the agent never gains access to the credentials. The company notes that handing an agent a password or secret is a common but risky path: a copy of the credentials is retained and remains outside the user's control. According to 1Password, the new service provides a 'trusted way to delegate tasks to personal AI agents while keeping control over passwords, secrets and credentials'.

How it works

The system builds communication between the agent, the user and the password manager. When an agent attempts to log into an account on the user's behalf, a unique digital key is sent to 1Password confirming the agent's digital identity. The user receives a notification about the login attempt and can personally approve the request. After approval, 1Password itself performs the login for the agent, filling in the credentials so that it can complete the task.

Supported agents

Users can choose an agent to their liking: Claude Duet from Anthropic, Dots or ChatGPT Work from OpenAI, Grok Bot, Muse from Meta, as well as agents from the startup Town. The company says it is working with other companies to expand the list.

Context: banks detect agent activity

The launch took place on the same day Axios reported that US banks had begun detecting agent activity inside users' accounts. This is not about hacks, but about a growing practice in which consumers entrust automated services to act on their behalf, including in financial matters. The problem is that most banks cannot determine whether such activity is linked to actual customer requests. Bankers interviewed by the outlet described the situation as 'horrifying'. A Chase representative, in a comment to Axios, added that customers 'should not share their usernames and passwords with anyone'.