1Password Allows Bots to Connect Without Exposing Credentials

1Password launches "1Password for Agents," an option that allows bots to connect to services without ever seeing the passwords. Approval occurs on the user's device, and agent identification is done through a unique key. Compatibility already covers several AI agents, with more integrations announced.

Initial Compatibility and Upcoming Integrations

The feature is announced as compatible with several AI agents: Meta Muse, ChatGPT Work and Dots, Claude Duet, Grok Bot, and Town. 1Password also indicates that partners such as Browserbase, Browser Use, Dana, Instinct, Kernel, Perplexity, Ramp, Tomo, and Wajo are working on integrating this functionality, with additional collaborations planned.

Local Approval and No Unlimited Access

Agents do not gain permanent access to accounts: each connection attempt must be approved by the user. The application sends a notification specifying the required credentials and the reason for the request; if approved, 1Password fills out the login form without transmitting the secrets to the agent. Approval occurs on the user's device, a process described as similar to two-factor authentication. Once connected, the agent can continue its task without having accessed the password.

Identity Key for Agents and Management's Cap Set

The new "1Password for Agents" option allows the execution of tasks requiring a login without exposing the password and is presented as simplifying these uses. Technically, an agent authenticates with 1Password via a unique digital key when it reaches a login form. CEO Aashish Tripathi describes a context where agents are now performing concrete tasks for users, believing that the old model of sharing credentials is no longer suitable. He asserts that an agent should accomplish its mission without permanent access to a user's account resources and that 1Password provides the layer of trust enabling this functionality in personal use cases.