1Password launches secure agentic autofill with Browserbase so AI agents can access credentials without exposing them

Summary

1Password is launching what it calls secure agentic autofill, built in partnership with Browserbase, that lets AI agents request and use credentials stored in a user's 1Password vault without those credentials ever being exposed to the underlying model.

The problem CEO David Fogno is solving is specific: as agents become more capable, they increasingly need access to the same resources their human principals control — logins, API keys, sensitive form data — but the authorization frameworks built for deterministic software don't transfer cleanly to non-deterministic agents. The current workaround, which Fogno says he hears about constantly from founders across the AI ecosystem, is developers hard-coding credentials directly into agent workflows. That creates both a security exposure and a visibility gap that compounds as those agents scale.

The integration works by connecting a Browserbase-hosted agent directly to a user's 1Password vault. When the agent needs a credential mid-task, it requests access, the user authorizes it in real time, and the credential flows through end-to-end encryption — never sitting in an LLM's context window or a third-party log.

Browserbase partnership rationale

Fogno frames the Browserbase launch as a reference implementation rather than an exclusive relationship. 1Password built the integration to be platform-agnostic, and the stated intent is to be the credential layer inside any agent infrastructure — hyperscaler or otherwise. Browserbase served as the partner to get the design right before broader rollout.

Business snapshot

1Password serves 175,000 corporate customers, a segment that now represents nearly 80% of revenue. Fogno describes the company as profitable and growing, positioning it as an identity security solution for the enterprise rather than a consumer password manager. The agentic autofill capability is framed as the next leg of that enterprise identity story, not a pivot.

On the broader AI market, Fogno lands on a straightforward split: teams either hard-code credentials and accumulate hidden risk, or they lock down agent use entirely because security teams say no. Both outcomes are bad. The product bet is that removing the credential friction unlocks deployment that would otherwise stall.