Cloudflare just added OpenAI's GPT-5.4 and Codex models to Agent Cloud, its enterprise platform for building and running AI agents on Cloudflare's global network.
The integration gives companies access to two OpenAI models through Cloudflare's infrastructure:
- GPT-5.4 for general reasoning and multi-step agent tasks
- Codex for code generation and software automation (Codex is OpenAI's model trained specifically on programming languages and code repositories)
The pitch to enterprise buyers is consolidated control. Instead of stitching together a model API, a hosting layer, and security tooling from separate vendors, Agent Cloud bundles those concerns into one platform. For teams already using ChatGPT for individual work, this is the enterprise-grade step up - agents that take actions autonomously rather than just answer questions.
Cloudflare's edge network is a real differentiator for agentic workloads. When an agent needs to make a dozen API calls to complete a task, latency compounds fast. Running inference close to where data originates cuts that overhead, and for globally distributed apps or teams spread across regions, that's a genuine operational advantage, not just a spec sheet talking point.
The Codex integration is most relevant for development teams. Automated code review, test generation, and refactoring pipelines are the obvious use cases - the repetitive work that burns engineering hours. Teams building AI-assisted development workflows might pair this with dedicated coding tools like Claude Code depending on model preference and existing toolchain.
No public pricing is listed. Enterprise integrations like this typically involve custom contracts, which puts it out of reach for smaller teams without a formal procurement process.
The broader pattern: cloud infrastructure providers are embedding AI models directly into their platforms rather than just selling raw compute. AWS, Azure, and Google Cloud are all doing versions of this. Cloudflare's angle is its edge network - faster for latency-sensitive agent workflows, with security controls built in at the infrastructure level. The company detailed the partnership in a blog post published April 13.