1Password for Agents lets personal AI agents log in without ever seeing the password
Identity security company 1Password today launched 1Password for Agents, a new sign-in system for personal artificial intelligence (AI) agents.
Agents from Meta Platforms Inc., Anthropic PBC, OpenAI Group PBC and Elon Musk’s SpaceXSI will be able to use the system to sign in to people's accounts without ever seeing their passwords.
1Password for Agents is designed to solve a problem that grew as AI agents began paying bills and planning vacations for people. These tasks often require signing in to an account. The easiest way has been to paste the password into the agent's chat window. But this leaves a copy of the password stored somewhere the account owner cannot control.
Aashish Tripathi, chief product officer at 1Password, said agents are “starting to do real work on our behalf.” He believes that when an agent is asked to complete one task, it should not get permanent access to the entire account.
With the new system, the password never goes to the agent. Instead, a unique digital key tells 1Password which agent is trying to sign in. When the agent requests access, the request explains which account it wants to use and why. Nothing happens until the person approves the request on their own device. After approval, 1Password enters the password directly into the login page.
OpenAI has connected ChatGPT Work and its always-on Dots agents to 1Password. Meta's Muse requests login details from 1Password whenever the connection is turned on. Town.com Inc., a startup that makes work assistants and raised $108 million this week, has also joined. With 1Password connected, SpaceXSI's Grok Bot can sign in without anyone pasting a password into it, according to Tido Carriero, the company's head of product.
Claude has worked with 1Password in the browser since July. The next step is Anthropic's Claude Duet agent, released Wednesday. Duet asks for permission before signing in to any website. Anthropic Chief Product Officer Ami Vora said 1Password will “soon” enter passwords people have already set up to complete these sign-ins. Anthropic wants to make this kind of help available to people “who’ve never before let an agent near their accounts,” she said.
In addition to the companies already involved, nine more companies are building their own connections to 1Password. Perplexity AI Inc. and Browserbase Inc. are among them. Their browser-based agents have been connected to 1Password since late 2025.
Canva Pty Ltd. and Stripe Inc. are among the more than 200,000 businesses that use 1Password. The company did not say when each partner's connection will become available to users.
The launch follows an argument made last month by 1Password Chief Technology Officer Nancy Wang at Okta Inc.'s Oktane conference. She said access for AI agents should be linked to individual tasks. Speaking with theCUBE, SiliconANGLE Media's livestreaming studio, she described an agent as “probably both” a machine and a human. Under a system that grants access based on tasks, an agent should have to “prove that it actually finished its last task with the right level of permissions” before receiving more access, Wang said.