1Password has introduced 1Password for Agents, a platform designed to allow AI agents to authenticate into user accounts without accessing or storing actual passwords. The system uses unique digital keys to identify agents, requiring explicit user approval on their personal devices before 1Password automatically inputs credentials into the target login page.
The platform supports integrations with agents from major companies, including OpenAI, Meta Platforms, Anthropic, and SpaceX. Specifically, OpenAI has connected ChatGPT Work and its Dots agents, while Meta’s Muse and SpaceX’s Grok Bot have also integrated the service. Anthropic is preparing to implement the feature for its Claude Duet agent, following its initial browser-based collaboration with 1Password in July.
Beyond the initial partners, nine additional companies are developing integrations, including Perplexity AI and Browserbase, which have been testing connections since late 2025. Furthermore, the startup Town.com, which recently secured $108 million in funding, has joined the platform. 1Password currently serves over 200,000 business customers, including organizations like Canva and Stripe.
The initiative aligns with 1Password’s strategy of task-based access for AI, as outlined by CTO Nancy Wang. The company emphasizes that agents should verify completion of previous tasks with appropriate permissions before gaining further access. While the platform is now live, 1Password has not provided specific timelines for when each partner integration will be fully available to all users.
