Webernetes is an experimental project developed by ngrok that ports a functional subset of Kubernetes directly into the browser, executing completely on the client side without relying on any backend server infrastructure. Built primarily as an interactive simulator, it allows users to boot up isolated virtual clusters and define simulated container images directly in TypeScript / JavaScript. Instead of pulling heavy container binaries or running Docker daemons, Webernetes mimics core control plane mechanics, pod scheduling, and network handling down to HTTP routing, service discovery, and deployment strategies right in the browser runtime.
Here is a test project that I spun-up for Webernetes:
https://ndouglas-edera.github.io/my-webernetes-demo
When compared to spinning up a real infrastructure cluster on cloud providers or managed platforms, Webernetes eliminates the operational weight of traditional cluster management. A real production or staging Kubernetes deployment requires provisioning control nodes, configuring container runtimes like containerd, managing complex networking overlays, and continuously maintaining underlying compute instances. Webernetes bypasses this entirely by mocking container lifecycles and TCP/HTTP traffic natively in JavaScript. All of the instructions can be easily scripted in a 100 lines or less of TypeScript through a centralised main.ts source file:
https://github.com/ndouglas-edera/my-webernetes-demo/edit/main/src/main.ts
This translates to near-instantaneous cluster startup times, with zero setup friction for end users, and a radically reduced maintenance burden for educators, technical writers, or visual platform creators who want to showcase Kubernetes concepts without wrestling with real cloud instances. In my cases, I want to build cluster labs where users can perform insecure behaviours, container escapes, and other Kubernetes security test scenarios, without actually allowing those same users to abuse cloud resources through cryptominers, exfiltrate real credentials, or any other obviously undesirable outcomes.
From a financial perspective, the cost savings of moving simulation workloads to Webernetes are substantial. Hosting interactive tutorials, live documentation, or architecture demos on traditional cloud infrastructure incurs ongoing costs for control planes, compute nodes, load balancers, and egress bandwidth that accumulate regardless of traffic volume. Webernetes shifts the entire computational load to the visitor’s device, completely cutting out backend compute, hosting, and API costs. Because the browser executes the cluster logic locally, the hosting requirements collapse down to serving static web assets, making long-term interactive Kubernetes content practically free to maintain forever. Essentially, I can recreate a sort of "fake" Katacoda (now deprecated) or Instruqt lab that users can mess around with at any time.
Evaluating scale against a physical homelab setup, such as my existing cluster of Raspberry Pis, reveals a distinct shift in trade-offs between my existing physical resource limits and seemingly limitless browser-bound agility. While a physical Raspberry Pi homelab provides actual ARM nodes, real Linux kernels, and genuine network interfaces capable of running real Docker containers, it demands physical power, cooling, cabling, and ongoing hardware upkeep. Furthermore, a homelab cluster is bounded by fixed hardware capacity and local network constraints, making it tricky to share scale horizontally with thousands of concurrent online users. In contrast, Webernetes scales endlessly across web visitors because each user instantly gets their own isolated, sandboxed cluster powered by their local machine's memory and CPU. Although Webernetes cannot run real un-ported container images, mount true block storage, or handle real raw IP/UDP protocols like physical hardware can, its client-side architecture makes it far superior for scaling interactive learning experiences, rapid prototyping, and reproducible architectural demos to a global audience without ever burning out a single physical node.
If anything in this project sounds exciting, or if you have questions for me on my usage of this project, feel free to ask in the comment sections. I hope you're as excited about Webernetes as I am.