Identity security company 1Password has announced the launch of 1Password for Agents, a new sign-in platform designed for personal artificial intelligence agents.
This platform will enable agents from Meta Platforms Inc., Anthropic PBC, OpenAI Group PBC, and SpaceX’s SpaceXSI to access user accounts without ever seeing the associated passwords.
1Password for Agents was developed in response to a growing trend among agents who manage various tasks such as paying bills and organizing travel plans for clients. These activities generally require access to user accounts, often leading to the insecure practice of copying and pasting passwords into chat windows, which poses risks to account security.
Aashish Tripathi, Chief Product Officer at 1Password, emphasized the importance of limiting agents’ access. He noted, “Agents are starting to do real work on our behalf,” and argued that an agent tasked with performing one job should not retain sweeping access to an entire account.
With the new protocol, passwords are never shared directly with the agent. Instead, a unique digital key is used to authenticate the agent with 1Password when they reach a login page. The subsequent request specifies the account to access and the purpose of the request. The user must then approve this action on their device, at which point 1Password inputs the password directly into the login page.
The integration between 1Password and ChatGPT Work has been established, alongside the always-on Dots agents. Meta’s Muse agent also requests logins through 1Password when activated. Additionally, Town.com Inc., which recently secured $108 million in funding this week, has signed up for the service. SpaceXSI’s Grok Bot is now capable of signing in without requiring users to share their passwords, according to Tido Carriero, the head of product.
Since July, Anthropic’s Claude has been utilizing 1Password for browser access in the browser. The recently released Claude Duet agent, introduced on Wednesday, will also prompt the user for authorization before logging in anywhere. Anthropic’s Chief Product Officer, Ami Vora, stated that 1Password will soon autofill these logins by using passwords set up by users in advance. The goal is to make this technology accessible to individuals who have previously hesitated to allow an agent access to their accounts.
Beyond the initial launch partners, nine additional companies are in the process of developing their own integrations. Among them are Perplexity AI Inc. and Browserbase Inc., whose browser-based agents have been linked to 1Password since late 2025.
Canva Pty Ltd. and Stripe Inc. are part of the over 200,000 businesses currently utilizing 1Password. However, the company has not disclosed a timeline for when each partner’s integration will be accessible to users.
This announcement comes after 1Password Chief Technology Officer Nancy Wang advocated at Okta Inc.’s Oktane conference last month for linking agent access to specific tasks. During her discussion with theCUBE, SiliconANGLE Media’s livestreaming studio, she described an agent as “probably both” machine and human. She explained that under a task-based access model, an agent must “prove that it actually finished its last task with the right level of permissions” before gaining additional access.