1Password wants to stop people from pasting passwords into AI chat windows. On October 22 the company launched 1Password for Agents, a system that lets a personal AI agent sign in to websites with credentials stored in a user's vault, without the agent ever receiving the password or one-time code. The service is free for 1Password Individual and Family subscribers on 1Password.com.
The mechanics resemble a second-factor prompt. When an agent hits a sign-in page, it identifies itself to 1Password with its own digital key and asks for a specific login. The user gets a notification naming the account and the reason, and approves or rejects it on their own device. If approved, 1Password types the credentials into the destination page itself, then hands control back to the agent, which carries on with its task. According to The Verge, the approval step is mandatory for each sign-in, so connecting an agent does not open the whole vault.
The launch list reads like a roster of this autumn's agent apps. 1Password names Meta's Muse, Anthropic's Claude Duet, OpenAI's ChatGPT Work and Dots, Grok Bot from SpaceXSI, and Town's Townie assistant. Timing varies: the company says the Muse and Grok Bot integrations roll out over the coming days, while Claude Duet, the OpenAI products and Townie are listed as coming soon. Nine more companies, among them Perplexity, Browserbase, Browser Use, Kernel and Ramp, are building their own integrations.
Some of these links extend earlier work. SiliconANGLE reports that Claude has used 1Password in the browser since July, and that Perplexity's and Browserbase's browser agents have been connected since late 2025. What is new is a common approval model aimed at consumer and work agents that act on a person's behalf across many sites, rather than at developers wiring secrets into code.
The problem 1Password is targeting is real and growing. Agents that book trips, pay bills or file expenses stall at every login unless a human steps in, and the shortcut many users take is to give the agent the password directly, which leaves a copy outside their control and grants access well beyond the task at hand. Aashish Tripathi, an executive at 1Password, framed the product around that point: an agent should be able to finish the job it was given without gaining lasting access to everything behind the account. The company's chief technology officer, Nancy Wang, argued last month for tying agent permissions to individual tasks.
There are gaps. Accounts hosted on 1Password.eu and 1Password.ca are not supported yet, and 1Password says availability, platforms, safeguards and controls vary from one partner integration to the next. The design keeps the password away from the agent, but once it is signed in, the agent operates inside that account with whatever permissions the session carries. 1Password has not published an independent security review of the scheme.