Cloudflare expands Agent Cloud with new tools to build and scale AI agents

Cloudflare Inc. today announced an expansion of its Agent Cloud with new features that are designed to help developers build, deploy and scale agents.

The new release includes a suite of infrastructure, security and developer tools that help to move AI agents from experimental demos on local laptops to robust, production-grade workloads that run across Cloudflare’s global network.

As the industry is shifting toward coding agents and autonomous tools that read context, reason and take multi-step actions, existing infrastructure cannot scale in a world where every user and employee has dozens of personal agents running simultaneously. Cloudflare is aiming to address infrastructure, compute, deployment and security so developers can focus on building the next generation of applications.

“The way people build software is fundamentally changing. We are entering a world where agents are the ones writing and executing code,” said co-founder and Chief Executive Matthew Prince. “But agents need a home that is secure by default, scales to millions instantly and persists across long-running tasks. We’ve spent nine years building the foundation for this with Cloudflare Workers. Today, we are making Cloudflare the definitive platform for the agentic web.”

Cloudflare also highlighted support for leading artificial intelligence models, including those from OpenAI Group PBC, with the platform designed to run agents powered by GPT-5.4 and Codex for enterprise workloads at scale

Leading the list of new features is Dynamic Workers, an isolate-based runtime designed to run AI-generated code in a secure, sandboxed environment faster and more efficiently than traditional containers. The feature can spin up in milliseconds when an agent needs to execute a code snippet to call an application programming interface, transform data or chain tool calls together.

Key to Dynamic Workers is that they are also affordable, with Cloudflare saying that for the vast majority of agent tasks, the feature offers secure isolation at 100 times the speed and a fraction of the cost of containers and can scale to millions of concurrent executions with no warm-up.

The second new feature, Artifacts, is a Git-compatible storage primitive built specifically for the agents-first era. Artifacts allows developers to create tens of millions of repositories, fork from any remote source and provide agents with a permanent home for code and data that is accessible to any standard Git client.

Sandboxes, meanwhile, gives agents access to a full operating system when they need one. Sandboxes offers persistent, isolated Linux environments with a shell, a filesystem and background processes, where an agent can clone a repository, install Python packages, run builds and iterate with the same tight feedback loop a human developer gets.

Think, is a framework within the Agents software development kit that is designed for persistence and allows developers to build agents that support long-running, multistep tasks rather than just responding to single prompts.

The final new release sees Cloudflare using technology from its acquisition of AI deployment startup Replicate Inc. It offers an expanded AI model catalog that allows developers to choose from a range of state-of-the-art proprietary models, including those from OpenAI such as GPT-5.4, alongside open-source models through a dashboard. The feature makes switching between model providers as easy as changing a single line of code without the requirement of managing multiple vendors.

Photo: Cloudflare