OpenAI acquires Ona to integrate secure, persistent cloud environments for long-running AI agents.

What happened

OpenAI has announced plans to acquire Ona, a startup specializing in secure, persistent cloud environments. The acquisition is explicitly aimed at expanding OpenAI's Codex capabilities to support long-running AI agents across enterprise workflows.

Technical Details

Currently, interacting with code-generation models like Codex is largely a stateless, transactional process. Developers send a prompt, receive code, and must manually execute it in their own local or cloud environments. Ona provides the missing infrastructure layer: secure, persistent, and sandboxed cloud compute. By deeply integrating Ona’s technology, OpenAI can move beyond mere code generation to actual code execution and iteration within a dedicated stateful environment. This means an AI agent can write code, run it, read the logs, debug failures, and deploy services over hours or days without losing context or requiring continuous client-side orchestration.

Why it matters

From an engineering perspective, this is a massive leap toward true autonomous agentic workflows. The hardest part of building AI agents today isn't the reasoning capability of the LLM; it's the secure, scalable infrastructure required to let those models safely interact with file systems, databases, and APIs over extended periods. Ona gives OpenAI a native "runtime" for its models. For enterprise engineering teams, this could eliminate the need to build custom container orchestration just to safely run LLM-generated code. It bridges the gap between an AI assistant that suggests code and an AI developer that autonomously resolves tickets in a secure sandbox.

What to watch next

Keep an eye on how OpenAI exposes this new capability via their API. We should look for new endpoints supporting stateful sessions, background task execution, and secure credential management for agents. Additionally, monitor how existing cloud providers and developer platforms respond, as OpenAI is now stepping directly into the compute and runtime orchestration space.