1Password launched a tool that lets users authorize AI agents to access accounts without disclosing the underlying passwords or credentials to the agents.
1Password launched 1Password for Agents on October 22, 2026, offering a way for users to delegate account sign-ins to AI agents without giving the agents direct access to their passwords. The system coordinates between the user, the agent, and 1Password, which handles the credential entry after the user approves a login request.
When an agent tries to sign in, it sends a unique digital key to 1Password to identify itself. The user receives an alert and can confirm the request. If approved, 1Password supplies the necessary credentials to complete the login on the agent’s behalf, rather than exposing those credentials to the agent itself.
The company’s launch supports a range of agents, including Claude Duet from Anthropic, OpenAI’s Dots or ChatGPT Work, Grok Bot, Meta’s Muse, and agents from Town. 1Password said it is working with other companies to expand the service. The article does not specify pricing, supported account types, or a broader availability timeline beyond the launch announcement.
The feature addresses a practical security concern as people grant automated assistants permission to act in their personal accounts. Sharing a password directly can leave a credential accessible to software beyond the user’s control; the approval step and credential handling by 1Password are designed to avoid that direct handoff. Users still need to decide whether to authorize an agent’s specific login attempt.
The launch coincided with reporting that U.S. banks had begun detecting agent-driven activity in customer accounts. That activity may reflect customers using agents themselves, rather than unauthorized access, but banks may have difficulty determining whether an agent’s actions were requested by the account holder. 1Password’s approach places user confirmation in the sign-in flow, while leaving the user responsible for deciding which requests to approve.