OpenAI Acquires Ona (ex-Gitpod): Codex Gets a Cloud
OpenAI Acquires Ona (Formerly Gitpod): What It Means for Codex
OpenAI announced on June 11, 2026 that it will acquire Ona — the German cloud development company you may have known as Gitpod. The deal folds Ona’s secure cloud execution and orchestration platform into OpenAI’s Codex ecosystem, giving Codex agents real persistent cloud environments to run in.
Last verified: June 13, 2026
TL;DR
- What: OpenAI to acquire Ona (ex-Gitpod), announced June 11, 2026.
- Why: Codex needs production-grade cloud sandboxes for long-running AI coding agents.
- Scale: Codex hit 5M weekly users; competitive pressure from Claude Code + Claude Fable 5 (released June 9, 2026).
- Impact: Codex tasks will be able to run autonomously in cloud environments while the developer’s laptop is closed.
- Status: Subject to regulatory review; close date not announced.
Who is Ona?
Ona was originally founded as Gitpod in 2014 in Kiel, Germany. The company pioneered remote, browser-based development environments — essentially cloud machines preconfigured with a codebase, dependencies, and tooling, so a developer could open a repo and start coding in seconds without local setup.
The rebrand to Ona happened earlier in 2026 as the company pivoted from “remote dev environments for humans” to “secure cloud execution and orchestration for AI agents.” Same core technology — Firecracker-style microVMs and Kubernetes orchestration — repositioned for the AI agent era. Customers include enterprises that needed isolated, ephemeral cloud environments where AI-generated code could run safely.
Why OpenAI wanted Ona
Three reasons stand out:
1. Codex needed cloud-native execution
The original Codex CLI runs locally on a developer’s machine. The Codex desktop app and IDE plugins still rely heavily on the developer’s local environment plus short-lived OpenAI-hosted sandboxes. As Codex agents grew more autonomous — sub-agents, multi-step tasks, hours-long refactors — the local model started to break. You don’t want a multi-hour refactor to die because the developer closed their laptop.
Ona gives Codex persistent, isolated cloud environments. An agent can run for hours, spin up sub-agents, run tests, and report back when the developer reopens the app.
2. The Codex–Claude Code race intensified
Anthropic released Claude Fable 5 and Claude Mythos 5 on June 9, 2026, just two days before the Ona deal was announced. Claude Code leads on SWE-Bench Pro (Fable 5 at 80.3%) and ships deep platform integrations: sub-agents, hooks, skills, Agent SDK, managed cloud/web workflows. Codex needed a comparable cloud story, fast.
3. Enterprise sales required SOC 2-grade infrastructure
Ona had already done the enterprise hardening: isolation, audit logs, network policies, on-prem support. For OpenAI’s enterprise push (and the rumored ChatGPT Business + Codex bundle), buying Ona was faster than building.
What changes for Codex users
| Today (June 2026) | After Ona integration |
|---|---|
| Codex CLI runs locally; agents stop when terminal closes | Codex agents continue in cloud environments persistently |
| Short-lived OpenAI sandboxes for tool calls | Long-lived isolated environments with state |
| Limited parallel sub-agent execution | Orchestrated sub-agent fleets in Ona-style microVMs |
| Self-hosted execution requires user setup | First-party cloud sandbox, billed through OpenAI |
| Customer’s own cloud required for VPC isolation | Ona’s “run inside customer cloud” pattern likely available |
The TechCrunch/TNW framing — “Codex agents inside the customer’s own cloud” — matters for enterprises. Ona already supports running its control plane inside a customer’s AWS, GCP, or Azure account. If OpenAI keeps that capability, Codex becomes deployable in regulated environments (banking, healthcare, government).
How this compares to other agent sandboxes
| Platform | Owner | Use case | Cold start | Isolation |
|---|---|---|---|---|
| Ona (post-OpenAI) | OpenAI | Codex-first cloud agents | TBD | microVM + Kubernetes |
| E2B | E2B | Any agent framework | ~150ms | Firecracker microVM |
| Modal | Modal | Python + GPU agents | ~1–3s cold, ~200ms warm | Managed containers |
| Daytona | Daytona | Multi-tenant dev sandboxes | Varies | Container-based |
| Vercel Sandbox | Vercel | Generated-code execution | Sub-second | microVMs |
| Northflank | Northflank | Multi-tier sandboxes | Varies (Docker/gVisor/Firecracker) | Multi-tier |
| Cloudflare Sandbox | Cloudflare | Edge code execution | Sub-second | Isolates (V8) |
| Beam / Freestyle | Independent | Python/Node sandboxes | Sub-second | Various |
The difference Ona-as-part-of-OpenAI brings: first-party Codex integration. If you build on the Codex SDK, you’ll likely get Ona-backed environments by default with no separate signup. For cross-model agent builders (Claude + GPT + Gemini in the same app), independent sandboxes like E2B and Modal remain the neutral choice.
See: OpenAI Ona vs E2B vs Modal: AI agent sandbox platforms compared.
What’s not yet clear
- Pricing. Will Ona-backed Codex sessions count against ChatGPT Pro/Business credits, or be a separate add-on?
- Branding. Does Ona keep its name, become “Codex Cloud,” or fold silently into the Codex API?
- Open-source story. Gitpod historically had a self-hosted, open-source edition. Whether that survives the acquisition is unclear.
- Regulatory. German-headquartered company, US acquirer — expect at least a Bundeskartellamt review.
What changes for Ona / Gitpod customers
Ona’s own announcement says existing customers will be supported through the transition. The realistic expectation:
- Short term (3–6 months): Service continues as-is.
- Medium term (6–18 months): Non-Codex workloads likely get a wind-down or migration path.
- Long term: Ona becomes OpenAI’s Codex Cloud; standalone product fades.
If you’re a Gitpod customer not using OpenAI, this is the moment to evaluate Coder, Daytona, or self-hosted alternatives.
What to watch next
- GPT-5.6 + Ona at launch. Multiple leaks point to a June 2026 GPT-5.6 release. If GPT-5.6 ships with first-party Codex Cloud powered by Ona, the agentic coding race resets again. See GPT-5.6 leaked features.
- Anthropic’s response. Claude Cowork already exists; expect Anthropic to push managed Claude Code more aggressively in Q3 2026.
- Pricing wars. OpenAI is rumored to cut API prices alongside GPT-5.6. Bundled Ona credits could be the differentiator. See GPT-5.6 vs GPT-5.5 vs Claude Fable 5.
Related coverage
- OpenAI Ona vs E2B vs Modal: AI agent sandbox platforms
- GPT-5.6 leaked features: June 2026 release
- Claude Fable 5 vs Mythos 5 vs GPT-5.5: public launch
- Cursor Auto Router vs Claude Fable 5 vs Windsurf SWE-1.5
Bottom line
OpenAI buying Ona is the moment Codex stops being a clever local CLI and becomes a cloud-native agent platform. Ona’s microVM execution and orchestration give OpenAI the same kind of infrastructure Anthropic already had via Claude Cowork and the Agent SDK. Expect a Codex Cloud product within 6 months and a real squeeze on neutral sandbox vendors that depend on Codex traffic.
Sources: OpenAI announcement (June 11, 2026), Ona blog (Johannes Landgraf, June 11, 2026), Bloomberg, CNBC, TechCrunch via TNW, Built In, CometAPI (June 11–12, 2026).