OpenAI/Ona: Codex gets cloud workspaces

OpenAI plans to acquire Ona and thereby extend Codex with secure, persistent cloud workspaces. According to OpenAI, Codex would no longer be tied only to a local device or an active session, but could continue longer-running tasks in controlled environments.

For companies, this is an important signal: agents are not only becoming more capable. They need a place where work is executed in a traceable, bounded and auditable way.

What was actually announced

OpenAI describes Ona as a provider of secure cloud execution and orchestration. The technology is meant to give Codex environments in which agents can use tools, systems and context over longer periods. Among the named tasks are development, testing, debugging, application modernization and remediation of vulnerabilities.

Ona positions its own platform similarly: background agents should move from developer laptops into standardized cloud environments. On its website, secure workspaces, sandboxed environments, governance and runtime security take center stage. Ona also points to typical software lifecycle tasks such as summarizing CI failures, patching vulnerable dependencies or reviewing pull requests.

Important: according to OpenAI, the acquisition is still subject to customary closing conditions and regulatory approvals. Until closing, OpenAI and Ona remain separate companies.

Why this matters

Many agent pilots start as a local developer workflow: one tool, one terminal, one repository. For production use, that is rarely enough. What becomes decisive is where the agent runs, which credentials it receives, which systems it may reach, how actions are logged and when a human reviews.

These exact questions shift the focus from “Which model writes better code?” to “Which operating boundaries do we accept?” For CIOs and IT leaders, that is closer to platform architecture than to prompt engineering.

What DACH companies should review

For organizations in the DACH region, the approach is interesting, but not automatically unproblematic. Persistent agent workspaces touch on data protection, source code protection, supplier management and internal control systems. Before a rollout, you should clarify at least four points: permitted repositories, secrets management, logging, and review and approval points.

The guiding question is not “May agents work longer?” but: in which environment may they continue work without losing control, traceability and accountability?