1Password’s new platform lets supported agents request account logins for user approval, then fills credentials without revealing passwords to the agent.
1Password announced 1Password for Agents on October 22, introducing a system for supported personal AI agents to use account credentials without receiving or storing the passwords themselves. The launch brings password management into a growing challenge for agent software: granting access to online accounts while keeping the user in control.
When an agent reaches a sign-in form, it presents a unique digital key to 1Password. The user can review which credential is requested and why, then approve or reject the request on their own device. If approved, 1Password fills the login details directly into the site’s form rather than passing them to the AI agent; the password remains in the user’s vault.
The company named integrations involving Meta’s Muse, OpenAI’s ChatGPT Work and dots, SpaceXSI’s Grok Bot, and Town’s Townie. Anthropic’s Claude Duet is also listed as a partner, though The Next Web reported that its sign-ins would follow later. 1Password also named Browserbase, Browser Use, Perplexity and Ramp among additional companies working on integrations. The announcement did not give pricing or individual partner rollout dates.
The product addresses a security trade-off in agent workflows: agents may need to log in to complete tasks, but sharing a password directly can expose a credential beyond the user’s control. The approval step gives people a way to authorize access for a particular request without handing the agent the underlying secret. It does not mean every agent or website is supported, nor does it remove the need to assess what an approved agent can do after signing in.
1Password said the platform builds on earlier work on secure agentic autofill. Its launch signals growing coordination between password managers and AI-agent providers as those agents take on tasks that require access to personal accounts and services.