OpenAI Acquires Ona to Let Codex Agents Run for Days Without Stopping

On June 11, OpenAI announced plans to acquire Ona, the company that grew out of cloud development environment platform Gitpod. Financial terms were not disclosed. The entire Ona team will join OpenAI’s Codex division, filling what has become Codex’s most visible gap: AI coding agents that stop the moment a developer closes their laptop.

Codex Has 5 Million Users — and One Big Problem

Codex now serves more than 5 million weekly active users. Most of those sessions are short: fix a function, write a test, refactor a module. Tasks that close out in under five minutes. The problem surfaces when complexity grows. Multi-hour or multi-day jobs require an execution environment that stays alive independently of the developer’s local machine. Close the VPN or shut the lid, and the whole workflow breaks.

Ona fills that infrastructure gap. Its persistent cloud sandboxes decouple agent execution from the developer’s workstation entirely. Tasks run in the background; developers can check in at any point to review progress, give instructions, or approve outputs — without needing to maintain a live connection.

Enterprise Buyers Want Control, Not Just Capability

Ona’s platform goes beyond keeping tasks alive. It was built with enterprise security compliance in mind: cryptographic hashing to block malicious applications regardless of filename obfuscation, restricted access to credential storage paths, and outbound connection filtering to suspected malicious servers. Everything leaves an audit trail.

That design matters enormously in large-enterprise procurement. When a security team asks “where does this agent actually run and what can it touch,” Codex currently struggles to give a clean answer. Ona gives OpenAI that answer. For compliance-heavy industries — finance, healthcare, defense contracting — this kind of infrastructure-level control is the difference between a proof-of-concept and a signed contract.

OpenAI Is Building a Full Developer Platform

The acquisition fits a clear 2026 pattern. In March, OpenAI bought Astral, the company behind popular Python tooling like Ruff and uv, pulling the Python toolchain layer directly into the Codex ecosystem. Also in March, OpenAI acquired Promptfoo to add security testing capabilities. Ona is the third developer-tools acquisition in under four months.

Six total acquisitions through June 2026 puts OpenAI on pace to nearly match its entire 2025 total of eight. The direction is consistent: Codex is no longer just an AI coding assistant. OpenAI is building a full software development platform.

Anthropic’s Advantage

Claude Code has a meaningful head start in enterprise autonomous coding. Anthropic’s first profitable quarter came primarily from Claude Code’s explosive growth among engineering teams. After closing its $65 billion Series H at a $965 billion valuation — surpassing OpenAI’s last private round for the first time — Anthropic has both the capital and the developer trust to press the advantage.

OpenAI’s response is this: buy the infrastructure layer that makes long-running agents viable in enterprise settings, then leverage Codex’s 5 million user base as the starting point for conversion. Whether that translates into enterprise contract wins depends on execution speed, not just technology. The procurement cycle in large organizations runs on 6-12 month timelines; the window that Ona opens is real but not guaranteed.

If this was useful, subscribe to the newsletter for weekly AI PM insights and GenAI case studies.

---

Sources:

• SiliconANGLE: OpenAI acquires AI agent orchestration startup Ona
• Techzine: As Anthropic claims the enterprise, OpenAI fights back with Ona deal