1Password lets AI agents sign in without seeing your password
The password manager says agents can request approved logins while 1Password keeps plaintext passwords and one-time codes hidden.
1Password has announced a new platform that lets personal AI agents sign in to websites without seeing the user’s password. Called 1Password for Agents, it puts the credential inside the password manager’s autofill flow while leaving approval with the person who owns the account.
How 1Password for Agents works
According to 1Password, an agent can request a credential from the user’s vault when it reaches a login page. The user reviews and approves that request on a trusted device, then 1Password fills the credential directly into the destination site. The agent does not receive the plaintext password or one-time code.
The company says the approach is designed to keep access limited to the task the user has approved. People can choose which credentials an agent may use and can revoke or manage access through 1Password, rather than pasting a password into an AI prompt.
The launch includes integrations with Muse from Meta and SpaceXSI Grok Bot, with 1Password also naming Anthropic’s Claude Duet, OpenAI ChatGPT Work and dots as coming soon. Other integrations listed by the company include Instinct, Browserbase, Browser Use, Dana, KERNEL, Perplexity, Ramp, Tomo and Wajo.
TechRadar reports that eligible Individual and Family customers on 1Password.com hosting will receive the upgrade at no extra cost. The article also notes that 1Password.eu and 1Password.ca hosted plans are excluded, so availability depends on the account’s hosting region.
Why it matters
AI agents are becoming more useful precisely when they can carry out tasks across logged-in services. That creates a difficult trade-off: an agent needs a way through the login screen, but handing the model a permanent password gives it more access than the task may require. 1Password’s proposal keeps the approval step visible and puts the credential handling in the password manager instead.
It is not a magic solution to agent security. Users still need to approve the right requests, and the destination site, agent integration and hosting region all matter. The design does, however, remove one particularly risky habit: copying a password into a chat or agent prompt.
Our opinion
The important part of 1Password for Agents is the boundary it draws around credentials. An agent can do more, but it does not automatically get to read the secret that makes the task possible. That is a sensible direction for agentic software, provided the approval prompts remain clear and the integrations expose useful logs.
The rollout is still early, with several named integrations listed as coming soon and regional exclusions for some customers. Treat it as a promising security layer rather than proof that every AI agent is now safe to hand the keys to.