Cloudflare Workers has evolved far beyond its initial role as a direct-to-developer platform. Today, it powers multi-tenant applications where customers provide their own code—from AI-generated TypeScript to CI/CD pipelines. The latest advancement, Dynamic Workflows, extends this concept to durable execution. In this article, we break down ten essential points about this new capability, explaining how it bridges the gap between static, pre-deployed workflows and the need for per-tenant, per-session logic. Whether you're building a SaaS platform, an agent-based system, or a CI/CD product, this list will help you understand why Dynamic Workflows matter.
1. The Journey from Workers to Dynamic Workflows
Eight years ago, Cloudflare Workers launched as a simple, serverless platform for developers. Over time, it grew into a multi-tenant ecosystem where platforms can deploy code on behalf of their own customers. This evolution introduced the need for dynamic deployment—allowing runtime code to run in isolated environments. Dynamic Workflows represents the next logical step: bringing the same dynamic, per-tenant flexibility to long-running, durable execution.
2. What Dynamic Workers Already Gave Us
With the Dynamic Workers open beta, Cloudflare provided a clean primitive for compute: hand the runtime some code, and in milliseconds you get an isolated, sandboxed Worker on the same machine. This enabled platforms to run arbitrary tenant code without pre-deployment. But compute alone wasn't enough—storage and source control also needed to be dynamic.
3. Durable Object Facets: Storage That Follows the Tenant
Durable Object Facets extended the dynamic idea to storage. Each dynamically-loaded app can now have its own on-demand SQLite database, with the platform acting as a supervisor. This ensures that every tenant's state is isolated, scalable, and manageable—critical for multi-tenant SaaS and agent-based applications.
4. Artifacts: Versioned Filesystem for Every Session
Artifacts brought dynamic deployment to source control. It's a Git-native, versioned filesystem that can be created by the tens of millions—one per agent, one per session, or one per tenant. This means every running context has its own persistent, versioned workspace, enabling sophisticated workflows that need file history.
5. The Missing Piece: Durable Execution for Dynamic Code
Despite these advances, one gap remained: durable execution. Cloudflare Workflows allowed long-running processes like video transcoding or multi-stage billing, but only when the workflow code was part of your deployment—a single class bound to the application. For platforms where each tenant has unique workflow logic, this static model broke down. Enter Dynamic Workflows.
6. What Is a Dynamic Workflow?
A Dynamic Workflow is a durable execution that is not bound to a pre-deployed class. Instead, the workflow code can be supplied at runtime—just like Dynamic Workers. It turns a run(event, step) function into a program that survives failures, sleeps for hours, waits for external events, and resumes where it left off—but now each tenant can ship their own unique workflow.
7. Why Traditional Workflows Fall Short in Multi-Tenant Apps
In a traditional setup, your wrangler.jsonc contains a binding that points to a single workflow class. This works fine when you control all the code. But if you're building an app platform where AI writes TypeScript per tenant, or a CI/CD product where each repo defines its own pipeline, there is no single class to bind. The workflow must be different for every request.
8. Real-World Scenarios: AI, CI/CD, and Agent SDKs
Dynamic Workflows unlock several critical use cases. In AI applications, the model can generate custom TypeScript for each tenant, and that code runs as a durable workflow. For CI/CD, every repository can define its own pipeline as a workflow that the platform executes. In agent SDKs, each agent writes its own durable plan—and Dynamic Workflows make that plan resilient and restartable.
9. How Dynamic Workflows Bridge the Gap
Just as Dynamic Workers solved compute isolation and Durable Object Facets solved storage isolation, Dynamic Workflows solve execution isolation for long-running processes. The platform can now hand the runtime a workflow definition, get back an isolated durable instance, and supervise its execution. This completes the trio: dynamic compute, dynamic storage, and dynamic durable execution.
10. The Future of Multi-Tenant Applications on Cloudflare
With Dynamic Workflows, Cloudflare is enabling a new class of multi-tenant applications. Platforms can let customers define their own business logic not just for short requests or storage, but for complex, stateful processes that may run for days. This is a key step toward fully programmable infrastructure where every tenant, agent, or session gets its own tailored, durable execution environment.
Dynamic Workflows represent the culmination of Cloudflare's vision for dynamic deployment. By combining the flexibility of runtime code with the reliability of durable execution, they empower platforms to offload complex orchestration to their users while maintaining full control and isolation. As multi-tenant SaaS and AI-driven agents become the norm, Dynamic Workflows will be a foundational primitive.