OpenAI to Acquire Ona to Power Persistent AI Agents

OpenAI plans to buy Ona, a company that builds secure cloud workspaces for AI agents, the company said on June 11. The OpenAI Ona acquisition targets a problem that's become obvious as Codex takes on bigger jobs: agents can plan and reason through complex tasks, but someone still has to give them somewhere to keep running after the person who assigned the work logs off.

Codex usage has climbed fast this year. OpenAI says more than 5 million people now use Codex each week, up roughly 400 percent from earlier figures, as the tool grows beyond its original focus on software developers. People increasingly use it for research, analysis, and multi-step automation that unfolds over hours, not minutes.

An AI agent can handle the thinking side of that without much trouble. The harder problem is execution: where does the work actually run once a laptop closes and the person walks away? That's the gap Ona is built to close.

What Ona's Cloud Execution Infrastructure Adds to Codex

Ona's core business has been building cloud-based development environments that are secure and easy to reproduce, work that OpenAI says has already reached roughly 2 million developers, including some customers the two companies already share. Folding that cloud execution infrastructure into the Codex ecosystem gives agents a dedicated, persistent space, one that can hold onto context, credentials, and project state across a task that spans days, not just minutes.

For everyday users, the practical effect should be agents that keep working in the background. Start a research task or a code refactor, step away, and come back later to check progress, give new direction, or review what got done, all without keeping a device connected the whole time.

Why Enterprises Care About Where Agents Run

Scale changes the calculus. Once an organization moves from testing agents to running them across real workflows, where those agents execute and who controls that environment becomes as important as how capable the underlying model is. OpenAI frames Ona's approach as customer-controlled: agents would operate inside a company's own cloud setup, with OpenAI supplying the underlying models and the orchestration layer that ties everything together.

Those production concerns tend to be specific: what an agent is allowed to touch, how its access gets scoped down, what gets logged for audit purposes, and who signs off before changes go live. Ona co-founder and CEO Johannes Landgraf has framed the company's role around that idea, describing Ona as the building blocks agents need for enterprise work: trusted, customer-controlled cloud environments where work continues across devices.

What Happens Before and After the Deal Closes

The OpenAI Ona acquisition isn't finalized yet. OpenAI's announcement on the Ona acquisition describes the deal as subject to standard closing conditions, including regulatory approval, with both companies continuing to operate independently until that process is complete.

Once it does close, Ona's team is expected to join OpenAI's Codex group, with a focus on extending secure, persistent execution to more enterprise workflows, including running tests, fixing bugs, addressing security issues, and modernizing existing applications over longer stretches of time. For developers already relying on Codex day to day, nothing changes immediately, but the direction is clear: an assistant built to keep working long after you've stopped looking at it.