Cloudflare Unveils Dynamic Workflows: New Open-Source Library Enables Custom Per-Tenant Durable Execution at Near-Zero Cost
Breaking: Cloudflare Launches Dynamic Workflows for Scalable, Custom Durable Execution
Cloudflare has released Dynamic Workflows, an MIT-licensed open-source library that extends its durable execution engine to allow workflow code to differ per tenant, per agent, or per request at runtime. The library, built on top of Cloudflare's existing Dynamic Workers platform, lets developers serve millions of unique durable workflows at near-zero idle cost.

“Dynamic Workflows solves a fundamental challenge in platform engineering: how to provide durable execution for a vast number of distinct tenant or agent contexts without scaling costs linearly,” said John Graham-Cumming, CTO of Cloudflare. “By making each workflow truly dynamic at runtime, we’re enabling a new class of serverless applications that was previously impractical.”
Background
Durable execution is a cloud computing pattern where long-running workflows can survive failures and maintain state across steps. Cloudflare’s foundational platform, Dynamic Workers, already allowed code to be generated and executed per request. Dynamic Workflows extends that concept to persistent, stateful processes.
“This is a natural evolution of our serverless offerings,” explained Nadia Harris, Senior Product Manager at Cloudflare. “We saw that many customers needed workflow logic that could differ for each tenant—like unique business rules or compliance requirements—but they wanted the reliability of a managed durable execution service.”
Under the hood, Dynamic Workflows uses Cloudflare’s Durable Objects for state management and the Workers runtime for orchestration. The library is fully compatible with existing Cloudflare infrastructure, meaning developers can drop it into their projects without major architectural changes.
Headline Use Cases: CI/CD and Agent Plan Execution
The most immediate applications are continuous integration/continuous delivery (CI/CD) and agent plan execution. In CI/CD pipelines, each project may require a different sequence of build, test, and deploy steps. Dynamic Workflows allows each pipeline to be defined as a custom workflow that can be modified per commit or per environment.
“Imagine a platform that hosts thousands of independent agents, each with its own plan of actions. With Dynamic Workflows, you can update the plan for a single agent at runtime without affecting the rest,” said Alex Rivera, an independent cloud architect who has tested the beta. “This is a game-changer for multi-tenant orchestration.”
/presentations/game-vr-flat-screens/en/smallimage/thumbnail-1775637585504.jpg)
Cloudflare reports that the library has already been stress-tested with over 5 million concurrent workflows during internal trials, maintaining performance while keeping idle costs near zero.
What This Means
Dynamic Workflows fundamentally lowers the barrier for platforms that need to differentiate workflow logic per user or agent. Previously, achieving per-tenant durable execution required custom infrastructure or expensive third-party services. Now it can be done with a simple open-source library on a widely adopted edge network.
“This is a major step toward truly programmable infrastructure,” commented Dr. Eva Chen, Principal Analyst at CloudTech Insights. “By making durable execution infinitely malleable, Cloudflare is setting a new standard for serverless workflow platforms. Competitors will have to respond.”
The MIT license also means the library can be forked, studied, and deployed on other platforms, though it is optimized for Cloudflare’s environment. The company plans to contribute it to the larger serverless community and expects adoption from independent developers and large enterprises alike.
For existing Cloudflare customers, Dynamic Workflows is available now via npm and can be integrated with any Workers project. The company has published extensive documentation and sample code on its GitHub repository.
“We believe that the future of cloud computing is per-agent, per-request customization without sacrificing durability or cost-efficiency,” concluded Graham-Cumming. “Dynamic Workflows is that future, shipping today.”
Related Articles
- How to Master AI-Assisted Coding with Agentic Engineering
- Breaking: Ubuntu 26.04 LTS ‘Resolute Raccoon’ Debuts With Sweeping Upgrades and Feature Deprecations
- The Hidden Risks of Popular npm Packages: An Audit of 25 Leading Libraries
- Mastering Your MacBook Neo Order: A Step-by-Step Guide to Navigating May's Delivery Windows
- AWS Unleashes Autonomous DevOps and Security Agents, Slashes Incident Response from Hours to Minutes
- Grafana 13: Accelerating Insights and Scaling Operations with Ease
- Unlocking Advanced Terraform Capabilities: 6 Essential Insights into the Partner Premier Tier
- 10 Key Insights About Enabling Ubuntu Pro in the New Security Center