Agents

1Password Launches Platform to Secure AI Agent Authentication Without Sharing Passwords

October 26, 2026 · 2 min read

Also published in العربية, Norsk

black circuit board
Cristiano Firmani / Unsplash

1Password has introduced a new platform called 1Password for Agents designed to allow artificial intelligence agents to access online accounts required for their tasks without transmitting actual login credentials to the models. Announced on October 22, this initiative addresses the security risks associated with AI agents performing actions on behalf of users, such as preparing expense reports or managing administrative duties, which traditionally require sensitive authentication steps. The system enables agents to interact with web services while keeping passwords and secrets isolated within the secure storage of the password manager.

The core mechanism separates the authorization to use an account from the knowledge of the associated credentials. When an AI agent encounters a login page, it submits an access request to the 1Password manager using a unique digital key. The user then reviews and authorizes this request from their own device. Upon approval, the password manager directly populates the login fields on the target website, ensuring the artificial intelligence model never receives or stores the password. This approach also applies to one-time codes, aiming to eliminate the risks of credential leakage or reuse that occur when passwords are shared via instructions or manual entry.

Integration efforts are currently underway with a variety of artificial intelligence agents and platforms. The service is being deployed for use with Muse from Meta, Grok Bot, Claude Duet, ChatGPT Work, dots, and Townie, the assistant from Town. Other entities involved in these planned developments include Browserbase, Perplexity, and Ramp. The rollout schedule varies by partner, with some integrations already in progress while others are slated for future release. The functionality and access methods differ depending on the specific agent and its level of integration with the identity management system.

At launch, the offering is available at no additional cost to Individual and Family subscribers who hold accounts on 1Password.com. However, availability is subject to geographic and hosting restrictions. Accounts hosted on 1Password.eu and 1Password.ca are not yet eligible for the service. The company emphasizes that while protecting the password is essential, it does not guarantee that an operation is legitimate or that the agent will restrict itself to the requested task. Therefore, the system relies heavily on human approval for each access request and maintains a strict separation between the agent and authentication data.

The effectiveness of this security measure will depend on the robustness of the integrations, the ability to trace operations, and the capacity to revoke access when necessary. This development highlights a shift in identity management where artificial intelligence agents act as intermediaries, necessitating tighter controls over their permissions. By preventing the transmission of secrets to the model, the platform aims to secure the growing trend of autonomous agents executing complex workflows across multiple online services without compromising user credentials.