10 Essential Insights into Cloudflare's Dynamic Workflows: The Future of Multi-Tenant Durable Execution

Cloudflare's platform has evolved dramatically since its early days as a direct-to-developer service. Today, it powers complex multi-tenant applications where each customer brings their own logic—AI-generated code, unique pipelines, or autonomous agents. The new Dynamic Workflows bridge a critical gap: enabling durable execution that dynamically adapts to each tenant, agent, or session without pre-defined bindings. Here are ten key things to understand about this transformative capability.

1. The Evolution of Cloudflare Workers

Eight years ago, Cloudflare Workers launched as a simple compute platform for developers. Over time, it expanded into a rich ecosystem supporting multi-tenant scenarios: platforms where users describe what they want and AI writes the code, SaaS products where every customer's business logic is unique TypeScript, and CI/CD systems where each repository defines its own pipeline. This shift from monolithic deployments to dynamic, tenant-specific code required new primitives for compute, storage, and now, durable execution.

10 Essential Insights into Cloudflare's Dynamic Workflows: The Future of Multi-Tenant Durable Execution — Source: blog.cloudflare.com

2. Dynamic Workers: Compute on Demand

The Dynamic Workers open beta gave platforms a clean primitive for runtime compute. It allows handing the Workers runtime fresh code at runtime and getting back an isolated, sandboxed Worker on the same machine in single-digit milliseconds. This means every tenant or agent can have their own compute environment without pre-provisioning or shared state, enabling true multi-tenant isolation and agility.

3. Durable Object Facets: Dynamic Storage

Extending dynamic deployment to storage, Durable Object Facets let each dynamically‑loaded app have its own SQLite database that spins up on demand. The platform acts as a supervisor, managing lifecycle and access. This eliminates the need for centralized databases or per‑tenant schemas, giving each tenant an independent data store that scales with their workload.

4. Artifacts: Dynamic Source Control

Source control also becomes dynamic with Artifacts—a Git‑native, versioned filesystem that can be created in tens of millions. One per agent, one per session, one per tenant. This allows each entity to have its own code repository, enabling autonomous development and version tracking without shared conflicts.

5. Enter Dynamic Workflows

Dynamic Workflows bridge durable execution and dynamic deployment. They allow workflows to be defined per tenant, agent, or request—not as part of a static deployment. This solves the problem where each customer or agent needs a unique durable plan, just as Dynamic Workers solved it for compute and Facets for storage.

6. Understanding Durable Execution

Cloudflare Workflows is a durable execution engine that turns a run(event, step) function into a program where every step survives failures, can sleep for days, wait for external events, and resume exactly where it left off after isolate recycling. It's ideal for onboarding flows, video transcoding, multi‑stage billing, and long‑running agent loops. Workflows V2 supports up to 50,000 concurrent instances and 300 new instances per second, redesigned for the agentic era.

7. The Problem with Static Workflow Bindings

Traditionally, Workflows assumed workflow code is part of your deployment—a fixed binding to a single class. That works for single‑tenant apps but fails when you need per‑tenant workflows. For example, an app platform where AI generates TypeScript for each tenant, or a CI/CD product with per‑repository pipelines, requires dynamic bindings. Dynamic Workflows remove this limitation.

8. Use Case: AI‑Generated Workflows

Imagine an app platform where users describe what they want and AI writes the implementation. Each tenant gets a unique workflow, generated on the fly. Dynamic Workflows let that AI‑written code become a durable, fault‑tolerant pipeline without manual deployment. The platform remains in control while tenants enjoy custom logic.

9. Use Case: Multi‑Tenant CI/CD Pipelines

In CI/CD products, each repository defines its own pipeline. With Dynamic Workflows, the pipeline code can be dynamically loaded per repository, with full durability. Failures during builds, tests, or deploys are handled automatically, and pipelines resume where they left off—critical for large, multi‑stage processes.

10. Use Case: Agent‑Generated Durable Plans

Agents that write and run their own tools need durable execution for long‑running tasks. Dynamic Workflows allow each agent to define its own durable plan, survive crashes, and continue from the last step. This enables autonomous agents to manage complex, multi‑step operations without human intervention.

Dynamic Workflows complete the vision of a fully dynamic, multi‑tenant platform on Cloudflare. By combining compute, storage, source control, and now durable execution into a cohesive primitive, developers can build applications that scale to millions of tenants—each with custom logic, isolated resources, and automatic resilience. As the platform continues to evolve, the boundaries between platform and tenant code blur, enabling truly adaptive, agent‑driven systems.