Introduction: The Age of Agents Has Arrived
In early 2026, Cloudflare hosted its first Agents Week, a dedicated innovation sprint centered on the rapid rise of autonomous AI agents. Over the past year, these self-directed programs have transformed how professionals work: coding agents accelerate software delivery, support agents resolve customer tickets end-to-end, and research agents synthesize hundreds of sources in minutes. But the most striking shift is that users now run multiple agents in parallel, often around the clock. This new operational reality demands a fundamental rethinking of cloud infrastructure.
As Cloudflare CTO Dane Knecht and VP of Product Rita Kozlov noted in the event’s kickoff, the potential scale is staggering. If even a fraction of the world’s knowledge workers each run a handful of agents simultaneously, the cloud must support tens of millions of concurrent sessions. The traditional one-app-serves-many-users model simply cannot cope. Developers and businesses want to build agents, deploy them to users, and run them at massive scale—all without rebuilding the stack.
The Agentic Cloud Vision: Cloud 2.0
Cloudflare’s response is what they call Cloud 2.0—the agentic cloud. This infrastructure is purpose-built for a world where agents are the primary workload. It solves problems across the entire stack: compute that scales from full operating systems to lightweight isolates, baked-in security and identity, an agent toolbox of models and tools, a clear path from prototype to production, and a web that adapts to growing agent traffic. The foundation? The containerless, serverless Workers platform launched eight years ago, now evolved into a complete ecosystem. During Agents Week, Cloudflare unveiled the next wave of primitives designed specifically for this new paradigm.
Compute Layer: Where Agents Run
Sandboxes: Persistent, Isolated Environments
Not all agents need the same compute. Some require a full operating system to install packages and run terminal commands; most need something lightweight that starts in milliseconds and scales to millions. Cloudflare’s Sandboxes give AI agents a persistent, isolated environment—a real computer with a shell, filesystem, and background processes. This enables agents to run complex operations without interfering with one another, and they can be spun up or down dynamically as demand fluctuates.
Artifacts: Git-Compatible Versioned Storage
Agents also need a place to store and version the code and data they generate. Artifacts provide Git-compatible versioned storage built for agents. Developers can create tens of millions of repositories, fork from any remote Git source, and hand off a URL to any Git client. This bridges the gap between agent output and human collaboration, making it seamless to iterate on code created by AI.
Security and Identity: Built-In, Not Bolted On
Agents operate autonomously, often across multiple systems and services. That demands security and identity that are integral to the runtime, not add-ons. Cloudflare’s platform includes fine-grained access controls, credential management, and policy enforcement designed for agent workflows. For example, agents can be assigned identities that limit what files, APIs, or databases they can access, ensuring that even if one agent is compromised, the blast radius is contained. This zero-trust approach is extended to every agent process, making security transparent to developers while remaining robust.
The Agent Toolbox: Models, Tools, and Context
To do real work, agents need the right set of tools. Cloudflare’s agentic cloud provides a toolbox that includes access to leading large language models (LLMs), vector databases for context, search capabilities, and a wide range of APIs. This allows developers to equip their agents with exactly the capabilities they need—whether it’s browsing the web, calling external services, or reasoning over large datasets. The platform abstracts away the complexity of managing these resources, letting developers focus on agent logic.
From Prototype to Production: A Clear Path
One of the biggest challenges in agent development is moving from a working prototype to a reliable, scalable production app. Cloudflare addresses this with seamless deployment pipelines, monitoring, and debug tooling. Code generated by agents can be directly deployed to Workers, connected to Sandboxes, and versioned with Artifacts—all without leaving the platform. Built-in observability gives developers real-time insight into agent behavior, bottlenecks, and failures. This streamlined path reduces the time-to-value for agentic applications from weeks to hours.
Adapting the Web for the Agentic Era
As agents drive an ever-growing share of Internet traffic, the web itself needs to evolve. Cloudflare is pioneering protocols and standards that make it easier for agents to navigate and interact with web services. This includes structured data formats optimized for agent consumption, new authentication flows for agent-to-service calls, and caching strategies that account for bursty, parallel requests from thousands of agents. The result is a web that not only supports human users but also millions of autonomous digital workers, ensuring low latency and high reliability.
Conclusion: Infrastructure for a New Computing Model
Cloudflare’s Agents Week showcased a comprehensive vision: an agentic cloud that is secure, scalable, and developer-friendly. By rethinking every layer—from compute and storage to security and web interaction—Cloudflare is building the foundation for Cloud 2.0. Developers now have the primitives to create agents that are more powerful, more reliable, and more integrated than ever before. As the age of agents accelerates, this infrastructure will be critical to unlocking the next wave of productivity and innovation.