OpenAI Buys Ona to Let Codex Work While You Sleep

OpenAI is buying its way toward coding agents that do not stop when you close your laptop. On June 11 the company announced an agreement to acquire Ona, the cloud platform formerly known as Gitpod, and fold its technology into Codex, OpenAI's software-development assistant. The price was not disclosed, and the deal still needs regulatory approval.

Ona is not a household name, but its job is specific. Founded in 2019 as Gitpod, it rebuilt itself around AI agents and rebranded in late 2025. It provides pre-configured cloud environments with access controls, audit trails, and the guardrails an autonomous agent needs to run safely inside a company's own infrastructure. In plain terms, it is the sandbox where an AI coder can be let loose without being handed the keys to everything.

Agents that outlast the session

The point of the deal, as Bloomberg reported, is endurance. Today a Codex task is tied to the machine that launched it. With Ona's orchestration underneath, OpenAI wants Codex to take on jobs that run for hours or even days inside a customer's secure cloud, carrying on long after the engineer who started the task has logged off for the night. That is a meaningful shift from an assistant you supervise to a worker you delegate to.

The timing is not subtle. OpenAI paired the announcement with a usage figure: Codex now has more than five million weekly active users, up from three million in April. Coding has quietly become one of the most reliable revenue stories in the industry, and the rivalry for that ground is sharpening.

A fight for the enterprise

That rivalry runs straight through Anthropic, whose Claude models have become a favorite among developers and which has been steadily expanding its enterprise tooling. Buying Ona lets OpenAI compete on the part that enterprises actually scrutinize: not how clever the model sounds, but whether its agents can be trusted to run unattended inside corporate systems without leaking data or going off the rails.

There is something worth sitting with here. An agent churning through a codebase overnight, with no human watching, is exactly the kind of capability that makes IT departments nervous and finance departments hopeful. OpenAI is betting that the answer to that unease is better plumbing rather than a more cautious pitch. We have written before about how enterprise AI agents have struggled to show clear returns; deals like this are an attempt to close that gap by making the agents dependable enough to leave alone.