Cloudflare has released Dynamic Workflows, an MIT-licensed library that extends its durable execution engine to enable workflow code to vary per tenant, agent, or request at runtime. Built on top of the company's Dynamic Workers platform, the library allows organizations to serve millions of unique durable workflows with near-zero idle cost—a breakthrough for multi-tenant and per-agent code execution.
“Dynamic Workflows unlocks the ability to dynamically define workflow logic for each execution context without sacrificing durability or performance,” a Cloudflare spokesperson told press. “This is a game-changer for platforms that need to run customized business processes at scale while keeping infrastructure costs flat.”
The library's headline use cases include CI/CD pipelines and agent-based plan execution, where each tenant or agent requires distinct workflow steps. By making workflows truly dynamic at runtime, Cloudflare aims to eliminate the trade-off between customization and operational efficiency.
Background
Durable execution engines—like Cloudflare’s—allow code to survive failures by saving state and replaying steps. Traditional implementations, however, require static workflow definitions that are difficult to customize per tenant.
Dynamic Workflows builds on Cloudflare’s existing Dynamic Workers technology, which already provided per-request code differentiation. Now, the same runtime flexibility extends to long-running, stateful workflows, enabling platforms to shift from “one size fits all” to truly bespoke execution patterns.
The library is released under the permissive MIT license, encouraging broad adoption and contribution. Cloudflare also emphasized that the solution incurs “near-zero idle cost,” meaning platforms can maintain thousands of unique workflow templates without paying for unused compute resources.
/presentations/game-vr-flat-screens/en/smallimage/thumbnail-1775637585504.jpg)
What This Means
For platform teams, Dynamic Workflows removes a major bottleneck: the need to pre‑define all possible workflow variants. Instead, they can generate workflow logic on the fly based on tenant, agent, or request parameters, while still benefiting from Cloudflare’s built-in durability and reliability guarantees.
“CI/CD systems and agent orchestration frameworks are the perfect proof points,” noted an industry analyst not affiliated with Cloudflare. “Being able to define a unique workflow for every deployment or every autonomous agent without a separate infrastructure cost is a powerful architectural shift.”
Early adopters are already testing Dynamic Workflows for scenarios like multi-tenant SaaS backends, personalized data pipelines, and programmable agent runtimes. The library is available today on GitHub and integrates with Cloudflare’s durable objects and workers ecosystem.
What’s next? Cloudflare plans to add more templates, observability tooling, and performance optimizations in coming releases. For developers, the immediate takeaway is that custom workflows no longer mean custom infrastructure.
- Key features: per-tenant, per-agent runtime differentiation
- License: MIT, open source
- Cost model: near-zero idle cost