Cloudflare Integrates Claude Managed Agents with Sandboxes for Scalable, Secure Agent Workloads

Cloudflare and Anthropic have released a new integration that lets developers run Claude Managed Agents on Cloudflare’s sandboxed environment, combining the agent loop on Anthropic’s platform with code execution on Cloudflare. The integration is shipped as a default deployment template on GitHub, providing enhanced security through customizable proxies, sandbox metrics, logs, SSH access, and lightweight V8 isolate sandboxes that boot in milliseconds. Developers can choose between full stateful Linux microVMs via Cloudflare Sandboxes, dynamic workers, or the new isolate backend for cost‑effective scaling. The integration also supports private service connectivity using Cloudflare Mesh and Workers VPC, allowing agents to access internal services without exposing them to the Internet. Additional features include email support per agent, browser control with session recording, and the ability to add custom tools by deploying functions. The integration leverages Cloudflare’s Agents SDK, Browser Run, and Dynamic Workers to provide a programmable, observable, and secure environment for agent workloads. By deploying the template, teams can quickly spin up agents, monitor sandbox activity, and scale to tens of thousands of concurrent agents using isolates.

The new integration addresses the need for scalable, secure agent execution at Internet scale, offering a lightweight alternative to full microVMs and enabling developers to run complex workflows while maintaining strict security boundaries and observability.