Bridging Durable Execution and Dynamic Deployment with Dynamic Workflows

Introduction: The Evolution of Cloudflare Workers

Since the launch of Cloudflare Workers eight years ago, the platform has evolved dramatically. Initially a direct-to-developers tool, it has expanded into a robust ecosystem where platforms not only build on Workers but also empower their customers to ship code through multi-tenant applications. Today, we see a wide range of use cases: AI-generated implementations based on user descriptions, multi-tenant SaaS where each customer's business logic is runtime TypeScript unknown to the platform, agents that write and run their own tools, and CI/CD products where every repository defines its own pipeline. This evolution has driven the need for dynamic deployment across compute, storage, and source control.

The Need for Dynamic Execution

Last month, with the Dynamic Workers open beta, we gave platforms a clean primitive for compute: hand the Workers runtime some code at runtime, and get back an isolated, sandboxed Worker on the same machine in single-digit milliseconds. Durable Object Facets extended this idea to storage—each dynamically-loaded app can have its own SQLite database, spun up on demand, with the platform acting as a supervisor. Artifacts did the same for source control: a Git-native, versioned filesystem creatable by the tens of millions, one per agent, per session, per tenant. With dynamic deployment for storage and source control in place, one critical piece remained: durable execution.

Understanding Durable Execution with Cloudflare Workflows

Cloudflare Workflows is our durable execution engine. It transforms a run(event, step) function into a program where every step survives failures, can sleep for hours or days, waits for external events, and resumes exactly where it left off when the isolate is recycled. This makes it ideal for processes that must persist beyond a single request: onboarding flows, video transcoding pipelines, multi-stage billing, long-running agent loops, and more. With Workflows V2, the engine now supports up to 50,000 concurrent instances and 300 new instances per second per account, redesigned for the agentic era.

The Limitation: Tied to Deployment

However, Workflows has always had one baked-in assumption: the workflow code is part of your deployment. Your wrangler.jsonc includes a block that binds the engine to a single class called MyWorkflow. One binding, one class, per deploy. This works well when you own all the code and run a traditional application. But it breaks down the moment you want to let your customers ship their own workflows.

Real-World Scenarios Requiring Dynamic Workflows

Consider an app platform where AI writes TypeScript for every tenant. Or a CI/CD product where each repository has its own pipeline. Or an agent SDK where each agent writes its own durable plan. In all these cases, the workflow differs per tenant, per agent, per request. There is no single class to bind. This is the same challenge that Dynamic Workers solved for compute and Durable Object Facets solved for storage.

Introducing Dynamic Workflows: Bridging the Gap

Today, we are announcing Dynamic Workflows, which bridges durable execution and dynamic deployment. This new capability allows platforms to hand over workflow code at runtime—just like they can with compute and storage—while still benefiting from the resilience and long-running capabilities of Cloudflare Workflows. With Dynamic Workflows, each tenant, agent, or session can have its own distinct durable execution plan, created on demand and fully supervised by the platform. No more binding a single class per deploy. The result: a truly dynamic, multi-tenant durable execution environment.

What This Means for Platforms and Developers

Dynamic Workflows complete the picture of dynamic deployment on Cloudflare. Platform builders can now offer their tenants:

• Compute: Dynamic Workers for sandboxed, on-demand code execution.
• Storage: Durable Object Facets for per-tenant SQLite databases.
• Source Control: Artifacts for versioned, Git-native filesystems.
• Durable Execution: Dynamic Workflows for long-running, fault-tolerant processes—each tenant with its own logic.

This unlocks new possibilities: AI-written workflows that run reliably, multi-tenant pipelines that scale, and agent systems that persist across sessions. The platform acts as a supervisor, managing resources while giving tenants full flexibility.

Conclusion: The Future of Dynamic Platforms

Dynamic Workflows represent the next step in Cloudflare's vision for dynamic, multi-tenant platforms. By removing the deployment barrier for durable execution, we enable a new class of applications where business logic is truly tenant-specific, yet backed by world-class infrastructure. Whether you're building an AI agent platform, a CI/CD service, or a multi-tenant SaaS, Dynamic Workflows provide the durable execution primitive you need—without the deployment friction. Stay tuned for more details and examples as we continue to refine this capability.