OpenAI Just Turned Codex into an Agent Platform
Codex has 5 million weekly users and is growing 400%. The Ona acquisition gives it persistent cloud environments. changing it from a coding assistant into a platform for long-running agents.
TL;DR: OpenAI is acquiring Ona to give Codex persistent cloud environments. Codex now has 5 million weekly users (400% growth). The combination changes it from a coding assistant into an agent platform: one where agents run in customer-controlled cloud environments, persist across sessions, and handle workflows that span hours or days instead of minutes.
Key takeaways:
- Codex now has 5M+ weekly users, up 400%: it’s no longer just a developer tool
- Ona provides persistent, customer-controlled cloud environments for long-running agents
- This moves Codex from a coding assistant to an agent platform that competes with Cursor and Claude Code on infrastructure, not just model capability
- Enterprise governance is built in: customer-controlled execution, credential scoping, activity logging, work review
- The structural advantage is persistence: agents that don’t die when you close your laptop
Codex has 5 million weekly users. That’s up 400% from earlier this year. It’s no longer a tool for developers: it’s used for research, analysis, and automation across roles.
Here’s what that growth looks like in practice: the most valuable work on Codex now spans hours or days. Teams start an agent task, step away, and need the agent to keep running. They need to check in from a different device, review results, and provide direction: without keeping their original session open.
Codex couldn’t do that. Until today.
What Ona brings
Ona builds secure, cloud-based workspaces for agents. The key detail: these are customer-controlled environments. Agents run inside the customer’s cloud infrastructure, not on OpenAI’s servers. This matters because enterprise adoption of AI agents has been blocked by exactly this concern: where does the agent run, who controls the data, what happens when the laptop closes?
Ona’s architecture gives enterprises:
- Persistent execution: agents that keep running after you close your laptop. They run inside your cloud environment, not tied to any single device.
- Granular governance: control where agents run, what they access, how credentials are scoped. Full activity logging with work review workflows.
- Secure credential management: credentials stay in your environment, not OpenAI’s. The agent authenticates through your infrastructure.
- 2 million developers already use Ona’s platform, so this is proven infrastructure being folded into Codex, not experimental tech.
Ona’s co-founder and CEO Johannes Landraf put it directly: “Agents need more than intelligence; they need a trusted workspace.”
What this means for the coding agent landscape
The coding agent market has been a race on two dimensions: model capability (who has the smartest model) and agent loop quality (who handles context best). This acquisition adds a third dimension: infrastructure.
| Tool | Model | Agent Loop | Persistence |
|---|---|---|---|
| Codex + Ona | GPT-5.5 | Strong | Cloud-persistent: agents survive device disconnect |
| Claude Code | Claude Fable 5 | Strong | Local: dies with terminal session |
| Cursor Agent | Mixed models | Strong | Local: dies with IDE session |
The structural advantage of cloud-persistent agents isn’t theoretical. A code review agent that runs for 4 hours, a security audit that spans 10,000 files, a migration that takes 3 hours of compute: these are real workflows that currently require keeping a terminal or IDE open. With Ona, Codex agents can run them in your cloud, and you check the results when they’re done.
OpenAI’s Thibault Sottiaux (Core Products Lead) described the target: “Enterprises want powerful agents that can do real work while meeting the security and control requirements of their environments.”
How does Ona close the enterprise trust gap for Codex?
The biggest blocker for enterprise AI agent adoption isn’t model capability: it’s trust. Companies don’t deploy agents because they don’t know where the agent runs, who controls the data, or what happens if the agent goes rogue.
Ona’s customer-controlled model directly addresses this. Agents run in your cloud. Credentials stay in your environment. Activity is logged. Work can be reviewed before deployment. This is the kind of governance that lets enterprises move from “we’re experimenting with agents” to “we’re running agents in production.”
The timing matters. Gartner predicts 40% of agentic AI projects will be cancelled before 2027 due to unclear ROI and lack of governance. By building enterprise controls into Codex from the infrastructure layer up, OpenAI is positioning to capture exactly those deals: the ones that stall on trust, not capability.
What doesn’t change
Codex is still Codex. The Ona team joins OpenAI, but the tool you use today doesn’t change overnight. The acquisition needs regulatory approval. During the transition, both companies operate independently.
What changes is the trajectory. Codex is no longer just a coding assistant with a good model. It’s a platform for agents that run in your infrastructure, persist across sessions, and integrate with your governance model. That’s the kind of shift that changes which tools enterprises choose: not based on which model scores highest on benchmarks, but on which platform they can deploy in production.
The agent market was a model race. It just became an infrastructure race too.
FAQ
What does the Ona acquisition mean for Codex users? In the short term, nothing changes. The acquisition needs regulatory approval, and both companies operate independently during the transition. In the medium term, Codex agents will be able to run persistently in your cloud environment: meaning you can start a task, close your laptop, and check results later.
How does Codex’s 5M user count compare to competitors? Cursor is estimated at 1-2M users. Claude Code usage numbers aren’t public but Anthropic’s developer tools are growing fast. Codex’s 400% growth rate suggests it’s capturing the developer tools market aggressively, partly because it’s bundled with ChatGPT’s existing user base.
Is this a response to Cursor and Claude Code? Partially. OpenAI clearly sees the coding agent market as strategic. Codex is their most visible developer product. But the Ona acquisition is more about opening up enterprise use cases than competing on features. Cursor and Claude Code are strong coding assistants. Ona makes Codex a platform for production workflows, not just code generation.
Related Posts
- Best AI Coding Agents 2026. How Codex, Cursor, and Claude Code compare on capability
- Apple Just Entered the AI Agent Game, Here’s What Changed, How Apple’s Core AI and Xcode 27 change the agent landscape
- Cursor vs Claude Code vs Copilot: 6 months of daily use. Detailed comparison of the leading AI coding tools
OpenAI’s Ona acquisition announcement details Codex’s evolution into a persistent cloud environment for long-running agents.
OpenAI’s Ona announcement details Codex’s evolution into a persistent cloud platform for long-running agents.
This article was published on Agentic Up (https://agenticup.dev): practical guides for developers and founders building with AI agents. Reach me at hello@agenticup.dev.
