Cloudflare's platform has evolved significantly since the launch of Workers, expanding from a direct-to-developer service to a foundation for multi-tenant applications. Now, with the introduction of Dynamic Workflows, the company addresses a critical gap: enabling durable execution that is unique to each tenant, agent, or session. This article explores the journey, the challenges, and how Dynamic Workflows empowers platforms to run custom long-running processes safely and efficiently.
How did Cloudflare's Workers platform evolve to support multi-tenant applications?
When Workers first launched eight years ago, it was a straightforward platform for individual developers. Over time, Cloudflare expanded its ecosystem to allow platforms to build on Workers and even enable their customers to ship code through multi-tenant applications. Today, Workers power a wide variety of use cases: AI applications where users describe requirements and the system writes the implementation, multi-tenant SaaS where each customer runs custom TypeScript logic at runtime, agents that create and execute their own tools, and CI/CD products where each repository defines its own pipeline. This evolution set the stage for more dynamic and tenant-specific execution models.
What are the key components that enable dynamic deployment in Cloudflare's ecosystem?
Cloudflare introduced several primitives to support dynamic, per-tenant deployment. Dynamic Workers allow platforms to hand code to the Workers runtime at runtime, receiving an isolated, sandboxed Worker in single-digit milliseconds on the same machine. Durable Object Facets extend this concept to storage—each dynamically-loaded app can have its own SQLite database, spun up on demand, with the platform acting as a supervisor. Artifacts provide a Git-native, versioned filesystem that can be created in the tens of millions, one per agent, session, or tenant. These components together form a robust foundation for dynamic compute, storage, and source control.
What is the core problem that Dynamic Workflows solves?
Before Dynamic Workflows, Cloudflare's Workflows durable execution engine assumed that workflow code was part of a deployment—a single class bound per deploy via wrangler.jsonc. This works fine for traditional applications where you own all the code, but it fails when platforms need to let customers ship their own workflows. For example, an app platform where AI writes TypeScript for each tenant, a CI/CD product where each repository has its own pipeline, or an agent SDK where each agent creates its own durable plan. In all these cases, the workflow is different for every tenant, agent, or request, demanding a per-execution approach rather than a fixed class.
How does Dynamic Workflows bridge the gap between durable and dynamic execution?
Dynamic Workflows combines the durability of Workflows—where every step survives failures, sleeps for hours or days, waits for external events, and resumes exactly where it left off—with the dynamic deployment capabilities of Dynamic Workers. Instead of a single bound class, Dynamic Workflows allows the workflow code to be provided at runtime per tenant. This means each tenant can have its own unique long-running process, such as onboarding flows, video transcoding pipelines, multi-stage billing, or agent loops. The engine supports up to 50,000 concurrent instances and 300 new instances per second per account, redesigned for the agentic era.
What are the practical benefits of Dynamic Workflows for platform builders?
For platform builders, Dynamic Workflows eliminates the need to hardcode or statically bind workflow classes. They can now offer their customers the ability to define custom durable execution plans without compromising security or isolation. Each tenant gets a sandboxed, isolated worker with its own SQLite database (via Durable Object Facets) and its own versioned filesystem (via Artifacts). This allows for massive scale—tens of millions of workflows—while maintaining per-tenant logic. The result is a more flexible, scalable, and maintainable architecture for multi-tenant SaaS, AI agents, and CI/CD systems.
How does Dynamic Workflows interact with other Cloudflare primitives like Dynamic Workers and Artifacts?
Dynamic Workflows is the final piece in a trio of dynamic deployment primitives. As discussed, Dynamic Workers handle compute, Durable Object Facets handle storage, and Artifacts handle source control. Dynamic Workflows adds durable execution to this stack, enabling long-running processes that are tenant-specific. They work together seamlessly: a Dynamic Worker can trigger a Dynamic Workflow instance that uses a dedicated Durable Object Facet for its state and an Artifact for code or data. This cohesive integration allows platforms to build end-to-end solutions where every component is isolated, scalable, and on-demand.
What does the future hold for Dynamic Workflows in the context of agentic and AI-driven platforms?
As AI agents and autonomous systems become more prevalent, the need for per-agent durable execution will only grow. Dynamic Workflows is designed with this in mind, supporting high concurrency and dynamic code injection. Future developments may include tighter integration with AI workflows, where agents not only write their own code but also define their own durable plans. Cloudflare's approach of providing a unified dynamic stack—compute, storage, source control, and now durable execution—positions it as a key infrastructure provider for the next generation of intelligent, multi-tenant applications. The ability to run millions of unique, long-running processes safely will empower developers to create more sophisticated and personalized services.