OpenAI buys Ona to keep Codex agents running in the cloud

If you have ever asked an AI agent to tackle a big project, you know the frustration of having to keep your laptop open and your browser tab active just so it does not crash halfway through.

OpenAI wants to fix that. The company just announced its acquisition of Ona, a cloud execution and orchestration platform, specifically to give its Codex platform a secure and persistent runtime in the cloud.

Codex has been growing at a blistering pace. OpenAI shared that Codex now has over 5 million weekly active users, a massive 400% jump from earlier this year. What started as a coding assistant has expanded into broader knowledge work like research, deep data analysis, and end-to-end task automation.

Closing the laptop without killing the agent

As AI workloads get more ambitious, tasks no longer wrap up in two minutes. They take hours or even days to test, build, and debug.

OpenAI realized that tying an agent to a user's local machine or a fragile browser session severely bottlenecks what people can delegate. Ona solves this exact problem.

Ona has spent years moving local software workflows into the cloud, serving around 2 million developers with reproducible cloud environments. Integrating Ona means Codex agents will be able to keep working smoothly inside dedicated cloud environments even after you close your laptop and walk away.

You can simply check in from another device, give feedback, approve sensitive actions, and review the final output whenever you are ready.

Keeping control inside the enterprise firewall

For enterprise customers, handing agents keys to production databases and internal codebases is terrifying without ironclad governance.

This is where Ona's architecture becomes critical. It uses a customer-controlled execution model. OpenAI delivers the intelligence and orchestration, while the execution itself runs inside the customer's own cloud perimeter.

That setup allows IT teams to define exact credential boundaries, manage access policies, log every terminal command, and enforce human-in-the-loop approvals before changes go live.

The deal is currently subject to standard closing conditions and regulatory approvals, with both companies operating independently until the transaction officially closes. Once completed, the Ona team will directly join OpenAI's Codex unit.

Why it matters

We think this is one of OpenAI's smartest enterprise infrastructure plays to date.

Every AI lab is racing to build autonomous agents, but the real blocker in enterprise adoption has never been just raw model intelligence. It has been runtime reliability, compliance, and isolation. If an agent cannot safely access internal systems over a 48-hour window without leaking credentials or demanding an open browser window, it remains a toy.

By anchoring agent execution inside the customer's own infrastructure, OpenAI neatly addresses the paranoia of corporate security teams. It forces rivals to rethink their own runtime stacks, proving that the agent race is becoming an infrastructure battle as much as a model battle.

For knowledge workers and software teams, this points to a future where you genuinely hand off real work and pick up the results tomorrow. While you cannot deploy these new unified capabilities today due to pending regulatory reviews, this acquisition signals exactly where enterprise automation is heading next.