Cloudflare introduces new features to Agent Cloud to make AI agents more efficient, secure, and affordable for production environments. With features including Dynamic Workers, Git-compatible storage, and sandboxes, developers can deploy AI agents at scale and support long-running tasks.
The shift from simple chatbots to autonomous, context-aware AI agents requires powerful infrastructure that is both scalable and secure. Cloudflare is therefore adding new features to Agent Cloud, a platform that helps developers build and scale agents within a global infrastructure. This aims to address the limitations of traditional virtual servers and container-based environments, which are often too expensive and slow for mass adoption.
New features in Agent Cloud
Dynamic Workers form a core component of Agent Cloud. This isolate-based runtime starts up within milliseconds to execute AI code securely and efficiently. This is a hundred times faster than traditional containers and costs a fraction of the price.
Additionally, Cloudflare introduces Artifacts, a Git-compatible storage solution that scales for autonomous workloads. Developers can manage millions of repositories and provide agents with persistent code and data storage accessible via standard Git clients.
For agents requiring a full operating system environment, sandboxes are now generally available. These isolated Linux environments provide a shell, file system, and background processes, allowing agents to build, install, and iterate just like human developers.
With Think, a framework within the Agents SDK, developers can create agents that support long-running and multi-step processes. This solves the problem of short-lived agents that only respond to individual commands, enabling more complex workflows.