1Password introduces credential-free logins for AI agents
1Password has launched 1Password for Agents, a platform intended to let AI agents log into a user’s personal accounts without ever receiving that user’s passwords, secrets or other credentials. The service is designed around a three-way interaction between an AI agent, the user and the password manager.
When an agent needs to sign in, it sends a unique digital key to 1Password to confirm its digital identity. The user receives an alert about the login request and can approve it personally. Once approved, 1Password completes the sign-in by filling in the required credentials on the agent’s behalf, allowing the agent to carry out the requested task.
Delegation without handing over credentials
The product addresses a basic security problem in agent-driven workflows: giving an assistant a password or secret may be a quick way to automate a task, but it can leave sensitive information in copies that users no longer control. 1Password says its platform is intended to preserve user control while enabling task delegation.
1Password for Agents supports a range of named agent options, including Anthropic’s Claude Duet, OpenAI’s Dots and ChatGPT Work, Grok Bot, Meta’s Muse, and agents from the AI assistant startup Town. The company says it is working with other companies to extend the service.
Agent activity raises account-control questions
The release arrives as financial institutions are paying closer attention to automated activity within consumer accounts. Axios reported that US banks have begun detecting agentic activity, reflecting cases in which consumers allow automated services to act for them, including in financial matters.
The challenge for banks is determining whether an observed agent action was genuinely requested by the account holder. Bankers interviewed by Axios described that uncertainty as alarming, while a Chase spokesperson warned customers not to give usernames and passwords to anyone.
What businesses should take from the launch
1Password’s approach separates agent access from credential disclosure: the user remains in the approval loop, while the password manager performs the sensitive credential-entry step. That distinction may be important wherever agents need account access but must not retain reusable secrets.
For businesses evaluating AI-agent workflows, the practical implication is to favour designs that verify the agent, notify the account holder and require explicit approval before a password manager supplies credentials.