1Password Launches 1Password for Agents to Let AI Assistants Log In Without Revealing Passwords

Password manager 1Password has launched a new security tool that lets artificial intelligence assistants log into personal accounts without ever viewing the user’s actual credentials. Dubbed 1Password for Agents, the platform establishes a communication loop between the assistant, the user, and the password manager. When an assistant attempts a login, it sends a unique digital key to 1Password to verify its identity.

The user then receives an alert to confirm the request. Once approved, 1Password automatically fills in the credentials on the assistant’s behalf so it can complete the task. The tool currently supports various digital assistants, including Anthropic's Claude Duet, OpenAI's Dots and ChatGPT Work, Grok Bot, Meta's Muse, and startup Town.

The launch addresses growing security concerns as consumers increasingly trust automated software with sensitive personal information. 1Password warned in a blog post that handing over raw credentials to agents creates uncontrolled copies that expose users to risks. The tool arrives as U.S. banks reportedly detect growing automated activity inside user accounts.

While these actions often stem from consumer delegation rather than malicious attacks, financial institutions struggle to distinguish authorized agent traffic, prompting warnings from banks like Chase against sharing passwords with any third party.