OpenAI Acquires Ona to Run Codex Agents Inside the Customer’s Own Cloud

OpenAI said on Thursday that it has agreed to acquire Ona, the German cloud startup formerly known as Gitpod, in a deal designed to let its Codex coding agents keep working long after a developer has closed their laptop. Rather than folding Ona's product into ChatGPT, OpenAI is buying the thing that makes autonomous agents palatable to corporate IT: a secure, persistent place for them to run.

Ona builds cloud development environments that live inside a customer's own infrastructure. An agent spun up in one of these environments gets the tools, credentials, and context it needs to carry out a long task, then keeps running on the company's servers instead of OpenAI's. The roughly 79-person company has served around two million developers under the Gitpod name and rebranded to Ona in late 2025 as it pivoted toward agent workloads. "Agents need more than intelligence; they need a trusted workspace," co-founder Johannes Landgraf said.

That framing speaks directly to the anxiety holding back enterprise agent adoption. CIOs and CISOs have been wary of turning autonomous coders loose on production systems, where a confused agent can delete files, rack up runaway token bills, or take actions no one approved. Because Ona's environments run inside a customer-controlled virtual private cloud, organizations can wrap their own governance around them — credential management, access restrictions, and read/write protections. Jeremy Roberts of Info-Tech Research Group described the platform as "a bucket for the agents to operate in," where IT can "ensure that access is properly credentialed and controlled effectively to prevent the model doing what it shouldn't be doing."

The acquisition lands as Codex usage explodes. OpenAI says more than five million people now use the coding agent every week, up roughly 400 percent since the start of the year. Until now Codex has largely been tied to a single device; Ona's persistent sessions extend it to background work that survives across machines and retains memory between runs — closing a gap analysts had flagged in OpenAI's enterprise story.

The competitive subtext is hard to miss. Gartner characterized the move as OpenAI's answer to Anthropic supporting self-hosted sandboxes in its Claude Managed Agents, and several analysts read it as a bid to blunt Claude Code's growing traction with developers. "It feels like a move to keep pressure on Anthropic," said Tom Findling, chief executive of Conifers.ai. Both labs are now competing as much on trust and control as on raw model capability.

Terms were not disclosed, and the deal still needs regulatory approval before it closes; the two companies will continue to operate separately in the meantime. IDC pegged Ona's 2025 revenue at roughly seven million dollars, which on standard multiples would imply a price somewhere in the few-hundred-million range. It is OpenAI's second notable acquisition in a matter of months, following its March purchase of the AI security platform Promptfoo, and arrives as the company is reported to be preparing for an eventual public listing.