Cloudflare Brings OpenAI's GPT-5.4 to Agent Cloud for Enterprises

Millions of enterprise customers can now run OpenAI's frontier models, including GPT-5.4, directly inside Cloudflare's Agent Cloud, the two companies announced. The integration turns Cloudflare's edge network into a distribution channel for OpenAI's most advanced models and moves the competition among infrastructure providers from hosting raw AI APIs to powering production-grade autonomous agents.

Agent Cloud is Cloudflare's platform for deploying AI agents that perform real business work. With OpenAI models now available inside it, companies can deploy agents that automatically handle tasks such as responding to customers, updating internal systems, and generating reports — all within a secure, production-ready environment, according to the announcement.

The platform runs on top of Cloudflare Workers AI, the company's system for running AI models at the edge of its global network. That architecture matters for the enterprise market: it lets businesses build and deploy AI applications and agents that deliver fast, real-time experiences at global scale, rather than routing every inference request through a centralized data center.

The stakes: agents move from demos to production

The deal signals where both companies believe enterprise AI is heading. OpenAI has spent the past year building out what it calls the intelligence layer of the world's largest and most established enterprises. That effort now includes Accenture, Walmart, Intuit, Thermo Fisher, BNY, State Farm, Morgan Stanley, and BBVA, among others. The Cloudflare partnership extends the same strategy to a much broader base of customers — businesses that want frontier-model agents without building the underlying infrastructure themselves.

The scale numbers OpenAI disclosed alongside the announcement frame why distribution partnerships like this one matter. More than 1 million business customers around the world directly use OpenAI. Codex, the company's coding and agent harness, counts 3 million weekly active users. OpenAI's APIs process more than 15 billion tokens per minute.

For Cloudflare, the integration is part of a broader push to bring the most advanced AI capabilities to enterprises. The company is positioning its edge network as the substrate for what it calls the next-generation AI-native stack.

"By bringing OpenAI's powerful models directly into the Cloudflare environment, we are collapsing the distance between intelligence and the end user," said Dane Knecht, chief technology officer at Cloudflare. "This allows developers to build sophisticated, AI-driven applications and agents that aren't just smart, but are lightning-fast and globally scalable by default. Whether it's using Codex to streamline development workflows or deploying AI applications and agents at scale, we're proud to provide the infrastructure that makes the next generation AI-native stack possible."

Codex harness lands in Cloudflare Sandboxes

The announcement also covers OpenAI's Codex, and here the details get concrete. The Codex harness is now generally available in Cloudflare Sandboxes, a secure virtual environment where developers can build, run, and test their AI applications. Codex will also be available in Workers AI in the near future, according to the companies.

That matters because sandboxing is one of the hard problems in agentic AI. An agent that writes and executes code needs an isolated, controlled environment before an enterprise will let it anywhere near production systems. By making the Codex harness generally available in Sandboxes now — and in Workers AI soon — Cloudflare and OpenAI are giving developers a supported path from prototyping an agent to deploying it on the same edge infrastructure.

Enterprises using Agent Cloud can now deploy agents built on the Codex harness to Cloudflare. In practice, that means a developer can build an agent on OpenAI's tooling and ship it to Cloudflare's global network without re-architecting it for a different runtime.

"Cloud agents are quickly becoming a foundational building block for how work gets done, and with Cloudflare, we're making it dramatically easier for developers to deploy, production-ready agents powered by GPT-5.4 and Codex to run real enterprise workloads at scale," said Rohan Varma, product, Codex, at OpenAI.

The phrasing from both executives is telling. Knecht talks about infrastructure and the "AI-native stack." Varma talks about agents as a "foundational building block for how work gets done." Both frame the partnership not as a model-distribution deal but as plumbing for a shift in how software operates — from applications that respond to user input to agents that act on it.

Why the edge angle matters

Agent Cloud's dependence on Workers AI is the technical differentiator in this arrangement. Running inference at the edge — physically close to the user making the request — reduces latency. For customer-facing agents that respond in real time, that latency gap shapes the user experience. Cloudflare operates a global network designed for exactly this kind of distributed workload, and the company is betting that enterprises will prefer agents that run close to their users over agents that run in a handful of hyperscale regions.

The production-ready framing is doing work here too. Enterprises have been cautious about deploying agents beyond pilots, citing concerns about security, reliability, and control. Cloudflare's pitch is that its secure, production-grade environment — combined with Sandboxes for isolated testing — addresses those concerns directly.

A pattern, not a one-off

The Cloudflare integration follows a series of enterprise deals that OpenAI has struck with large, established companies. Walmart partnered with OpenAI to create what the retailer calls AI-first shopping experiences. Accenture, Intuit, Thermo Fisher, BNY, State Farm, Morgan Stanley, and BBVA have all announced collaborations that put OpenAI models into core business workflows. The common thread: OpenAI is embedding itself into the operations of companies that already run at massive scale, then extending the same capabilities downmarket through infrastructure partners like Cloudflare.

The numbers suggest the strategy is compounding. A million business customers, 3 million weekly Codex users, and 15 billion tokens processed per minute through the APIs indicate that enterprise demand for frontier models has moved well beyond the experiment phase.

What developers can do now

Developers and customers can start using OpenAI in Cloudflare through Cloudflare's developer documentation, which covers configuration through AI Gateway. The immediate capabilities available are GPT-5.4-powered agents in Agent Cloud and the Codex harness in Cloudflare Sandboxes, with Codex in Workers AI listed as coming in the near future.

The near-term question is adoption. Cloudflare has made frontier AI agents technically easy to deploy; whether enterprises move their agentic workloads from pilots to production on the edge network will determine whether Agent Cloud becomes a default path or another option in an increasingly crowded field of agent platforms. The Codex rollout to Workers AI, when it arrives, will be the next concrete signal of how deep the integration goes.