Cloudflare launches Claude agent environments with Anthropic

Cloudflare has launched Cloudflare Environments for Claude Managed Agents in collaboration with Anthropic, linking Anthropic's Claude platform with Cloudflare's network and developer tools.

The integration is designed to let organisations run core agent loops on Claude while using Cloudflare's infrastructure to execute code, connect to private services, and add task-specific tools. It is intended to support workplace AI agents that need tighter security controls and the ability to scale across large numbers of users.

At the centre of the product is a control plane based on Cloudflare Workers that creates isolated sandboxes for each agent session. Security and compliance controls are applied by default, giving businesses a more controlled environment for deploying autonomous systems.

The approach reflects a broader industry effort to make AI agents suitable for internal business use, particularly when software needs access to company systems or data. Providers have been trying to combine model intelligence with infrastructure that can enforce access rules, separate sessions, and keep records of what an agent does.

How it works

Developers using the service can choose between Linux-based microVMs for more complex tasks and V8 isolate-based sandboxes for workloads that need faster start times. This allows different types of agent workloads to run within the same framework.

The service also includes controls for outbound connections and support for private links to internal services through Cloudflare Mesh. Those connections can use post-quantum-ready encryption to help keep sensitive data off the public internet.

Another part of the offering is built-in observability, including audit trails and session recordings. This is intended to give organisations a way to review agent activity and support internal governance or regulatory requirements where records of automated actions are needed.

Developers can also extend agents by writing functions directly on Cloudflare's platform. In practice, this means an agent can be given access to custom tools within the same environment where its tasks are executed.

Partnership focus

The launch builds on the existing relationship between Cloudflare and Anthropic, which the companies describe as focused on the infrastructure needed to connect AI systems with real-world services and data. The product positions Cloudflare as an infrastructure provider for AI agents, while Anthropic supplies the model layer through Claude Managed Agents.

For Cloudflare, the move broadens the role of its Workers platform beyond web applications and edge computing into AI operations. It is seeking to place its network and security products at the centre of how businesses deploy agent-based software, particularly where those systems need to operate across distributed environments.

Competition in this part of the market has intensified as cloud providers, model developers, and security firms all try to define the stack for agent deployment. Buyers are increasingly looking not only at model performance, but also at containment, access management, and auditability, especially in sectors where automated tools may interact with private systems.

Cloudflare Co-Founder and Chief Executive Officer Matthew Prince said the product was aimed at the gap between model intelligence and operational controls. "The true potential of AI agents will only be realized when they can safely interact with the real world at massive scale," he said.

"By partnering with Anthropic, we are giving developers the ability to execute code and access private data securely-backed by the speed and scale of our global network," Prince added. "Now, businesses can focus on building the most innovative AI applications on the planet."