1Password now lets Claude log into websites for you without ever seeing your actual password

Started by TeaSpiller, Jul 17, 2026, 11:04 AM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

Topic: 1Password now lets Claude log into websites for you without ever seeing your actual password   Views(Read 45 times)
Active members in this topic:
TeaSpiller(1)

TeaSpiller

1Password added official support for Claude today, letting Anthropic's AI access saved credentials to complete browser based tasks without the actual password ever being exposed to the model. The passwords never reach Claude's context window, its memory, or Anthropic's own systems at any point in the process

Here's how it actually works, when Claude needs to sign into a website to complete a task, 1Password shows the user exactly which credential Claude wants to use and why, and only after the user approves does 1Password inject the credential directly into the page itself. Access is scoped tightly to that one task and ends the moment the task is finished, and 1Password says it checks afterward to make sure no secrets ended up exposed anywhere on the page during the process. Credit cards and saved identities aren't supported yet, so for now Claude's access is limited specifically to logins and one time passcodes

Alongside this, 1Password's browser extension is getting a new Agentic Mode that gives users direct control over browser based AI agents more broadly. When an AI agent takes over browsing, the extension locks down automatically, hiding the password interface entirely so agents can only use logins and one time codes with explicit case by case approval, and this protection applies even for people who haven't set up the Claude integration specifically

The feature requires a Mac, the 1Password desktop app and browser extension, and a Pro, Max, Team or Enterprise Claude plan alongside the Claude desktop app and Claude in Chrome extension. It works across any site where Claude in Chrome can already complete actions. Reaction in 1Password's own community has been sharply split, several early commenters called it a security nightmare and said flatly they'd never hand credentials to an AI agent no matter how the access is scoped, while others treated it as simply the practical next step now that browser agents are already completing real tasks on people's behalf regardless
// TODO: write better signature

Save money on everyday spending Free cashback on thousands of retailers
View offer