OpenAI Acquires Ona to Enable Persistent Agent Execution Beyond Single Sessions
OpenAI is acquiring cloud-execution platform Ona to extend Codex agents' ability to work across hours or days in secure, customer-controlled cloud environments.
Last verified:
Bottom Line
According to the OpenAI Blog, OpenAI is acquiring Ona, a secure cloud execution platform, to extend Codex agents’ ability to operate persistently across extended sessions—hours or days—within enterprise cloud environments. The deal, subject to regulatory approval, aims to bridge the gap between agent experimentation and production-scale deployment by embedding Ona’s customer-controlled infrastructure model into Codex’s orchestration layer.
Codex’s Shift From Session-Bound to Persistent Work
Codex, OpenAI’s application-layer platform, has grown to 5 million weekly active users—a 400% increase since earlier in 2026—as it expanded beyond software developers to support research, analysis, and automation workflows. According to the OpenAI Blog, the majority of Codex’s emerging value now emerges over extended timeframes rather than real-time interactions. Users increasingly need agents to continue operating after they close their laptops, check progress asynchronously, and integrate decisions made across days or hours—a pattern that single-session, device-bound tools cannot support.
Ona’s core competency directly addresses this gap. The company spent five years building infrastructure for developers to migrate software development from local machines into persistent cloud environments, accumulating experience with 2 million developers across reproducible, secure execution models. This operational track record positions Ona to accelerate Codex’s transition from conversational tool to agent-based production system.
Enterprise Scaling and Customer-Controlled Security
The acquisition announcement emphasizes Ona’s ability to address organizational governance concerns—a critical blocker for agentic AI deployment at scale. According to the OpenAI Blog, as enterprises move from agent pilots to production workflows, “capable models are only one part of what they need.” Organizations require assurance that agents operate within environments meeting their security, data residency, credential scoping, and audit-logging requirements.
Ona’s model allows agents to execute inside a customer’s cloud environment while OpenAI supplies the model intelligence and orchestration—effectively separating the control plane from the execution plane. This architecture preserves organizational data sovereignty without requiring customers to make separate infrastructure investments.
Timeline and Regulatory Status
The OpenAI Blog confirms the acquisition is subject to customary closing conditions, including regulatory approvals, but does not specify a timeline. Until closing, both companies operate independently. Post-closing, Ona’s team will integrate with OpenAI’s Codex organization to build out enterprise execution capabilities.
Why This Matters
This acquisition signals OpenAI’s shift from agent experimentation to agent-in-production economics. For enterprises evaluating agentic AI platforms, the addition of customer-controlled cloud execution removes a major deployment friction point—concerns about data residency and compliance governance. For competitors offering agents (Anthropic, Google), the deal underscores that orchestration and execution infrastructure are now table-stakes for capturing enterprise adoption. Teams planning multi-day agent workflows or considering Codex-based automation should expect a more mature enterprise deployment story post-integration, though integration timelines and feature parity remain to be announced.
Frequently Asked Questions
Why is OpenAI acquiring Ona instead of building this capability internally?
Ona brings 5+ years of production experience helping 2 million developers work in secure, reproducible cloud environments. The acquisition accelerates OpenAI's ability to add enterprise-grade security, governance, and multi-session orchestration to Codex without starting from scratch.
Will Ona's existing customers be affected?
The OpenAI Blog post does not specify the transition plan for Ona's 2 million existing developers. Typically, such acquisitions involve a transition period where the acquired product is integrated or sunset; further clarity is expected post-closing.
What does 'customer-controlled execution' mean in this context?
It means agents can run inside the customer's own cloud infrastructure rather than OpenAI's, giving organizations control over data residency, credential scoping, logging, and security boundaries while OpenAI provides the model intelligence and orchestration layer.