OpenAI to Acquire Ona to Power Persistent AI Agents

OpenAI announced in June 2026 that it will acquire Ona, a cloud infrastructure company whose technology keeps AI agents running across sessions, even when the device that started the work goes offline. The OpenAI Ona acquisition is built to push Codex past single-session work, letting agents carry tasks that stretch over hours or days without needing your laptop awake. For anyone who has watched a long job die the moment they closed the lid, that is the whole point.

Codex already reaches more than 5 million users each week, up 400% since earlier this year. What began as a developer tool has grown into a broader Codex ecosystem for research, analysis, and workflow automation. The more these agents take on, the more they need somewhere stable to actually do the work.

Why Persistent Agents Need More Than a Model

Long-running tasks don't fit neatly inside one session. An agent refactoring a large codebase or working through a multi-day research request has to hold context, reach its tools, and keep moving even when nobody is watching over its shoulder.

Ona has spent years on a version of that problem. The company helped 2 million developers move software development off local machines and into secure, reproducible cloud environments that keep running on their own. That experience lines up almost exactly with what Codex needs now.

"Agents need more than intelligence; they need a trusted workspace," said Johannes Landgraf, co-founder and CEO of Ona.

Ona's Execution Model Gives Enterprises Infrastructure Control

Here is where the deal gets practical for large companies. Instead of routing agent activity through OpenAI's own infrastructure, Ona's model runs agents inside the customer's own cloud. OpenAI supplies the intelligence and orchestration, while the organization keeps control over data boundaries, how credentials are scoped, what gets logged, and how finished work is reviewed.

Thibault Sottiaux, Core Products Lead at OpenAI, said the goal is to make Codex deployable for enterprises at the highest standards of security, governance, and operational trust.

For a security or platform team, that control is the difference between a stalled pilot and a real rollout. Model quality answers one question. Whether an agent can operate inside your own compliance rules answers a very different one, and that is the gap Ona was built to close.

What Comes Next as the Deal Moves Toward Closing

The OpenAI Ona acquisition still has to clear regulatory approval and the usual closing conditions. Until it does, the two companies will keep operating as separate and independent businesses.

Once it closes, the Ona team joins OpenAI's Codex group to expand secure, persistent execution across the software lifecycle, from running tests and resolving issues to application modernization and vulnerability management. For teams deploying persistent AI agents, that roadmap points toward agents that can own longer slices of real engineering work.

The signal underneath the deal is clear. OpenAI sees the next limit on Codex adoption sitting not in how smart the model is, but in the infrastructure enterprises need to trust agents at scale.