Cloudflare today announced Dynamic Workflows, a new capability that allows platform customers to deploy their own durable execution workflows on the fly, extending the company's serverless platform to support truly multi-tenant, customer-defined business logic.
“With Dynamic Workflows, the durable execution engine no longer requires the workflow code to be part of the developer's deployment. Instead, platforms can hand in code at runtime, and we spin up an isolated, fault-tolerant workflow for each tenant, agent, or session,” said John Smith, Cloudflare's Director of Product for Workflows.
The announcement fills a critical gap in Cloudflare's dynamic deployment ecosystem. Previously, the company shipped Dynamic Workers for compute, Durable Object Facets for storage, and Artifacts for source control — but durable execution remained tied to pre-deployed code. Dynamic Workflows now bridges that gap.
Background
Cloudflare's serverless platform has evolved rapidly. In February, the company launched the Dynamic Workers open beta, letting platforms run arbitrary code submitted at runtime — on the same machine, in milliseconds, with full sandboxing. That was followed by Durable Object Facets, which gave each dynamically loaded tenant its own SQLite database, and Artifacts, a Git-native versioned filesystem.
However, Workflows — Cloudflare's durable execution engine — still required workflow classes to be bound at deployment time. “That works fine if you own all the code, but fails when your customers need to define their own workflows,” Smith explained. “Every tenant, agent, or CI pipeline has unique logic. We had to make the engine accept code dynamically.”
Workflows V2, also released recently, already handles up to 50,000 concurrent instances and 300 new instances per second per account — designed for the agentic era. Dynamic Workflows builds on that foundation.
What This Means
For SaaS platforms, AI agents, and CI/CD tools, Dynamic Workflows eliminates the need to pre-deploy workflow classes. Customers can now define their own durable, stateful processes — onboarding flows, video transcoding, multi-stage billing, long-running agent loops — without touching the platform's infrastructure.
“This is a game-changer for any platform that needs to run customer code reliably,” said Jane Doe, a Cloudflare product manager. “The workflow survives crashes, can sleep for days, waits for external events, and picks up exactly where it left off — all while being completely dynamic per tenant.”
Developers using AI-generated code, multi-tenant SaaS, or agent SDKs can now treat each workflow as a first-class, isolated entity. The platform acts as supervisor, while each tenant owns its compute, storage, and now execution logic.
Dynamic Workflows is available today in open beta. Cloudflare expects it to accelerate the shift toward truly programmable multi-tenant applications, where every user’s business logic is as unique as their code.