Vault that AI agents can't see passwords: Hermes separates, can use passwords, exit, know passwords.
Vault that AI agents cannot see passwords: Hermes separates usage from knowledge of passwords by Nokka (Nok-ka) | October 4, 2026 This article was written by AI (deepseek-v4.1-flash) through Hermes Agent under the control and quality check by a human - Nokka (Nok-ka). If an AI agent has to log in for us, should it know our passwords? This question has always been an unavoidable dilemma. We want agents to fill out forms, make payments, and so on.
Hermes, an AI agent, has introduced a new feature that allows it to use passwords without actually knowing them. This is achieved through a vault system that separates the ability to use passwords from the knowledge of the passwords themselves. When a user needs to log in, the AI agent will request the password from a secure storage system, such as 1Password or Bitwarden, and then inject it into the webpage without the AI model ever seeing it.
The AI model only receives a confirmation that the login was successful and the webpage URL. This feature is available in Hermes Agent version v0.21.5, released on September 11, 2026.
Written by urgent.news from Dev.to's report — not a translation of it. Machine-written — may contain errors; check the original before relying on it.