Kubernetes deploy is often associated with Helm charts and lots of YAML. But there are cases when you want to roll out an app faster with less manual setup. Many teams find: for a typical web service or microservice, a full Helm chart is overkill. Here's when you can skip Helm and how to do it.
Why Skip Helm
Helm fits well for replicating one chart across dozens of envs (dev, staging, prod, feature branches) and managing dependencies. But for a simple app — one container, a few replicas, one port — it adds abstraction layers (templates, values, hooks) that aren't always needed.
Writing and maintaining a whole chart for one app isn't always justified. Especially for small teams and infra that doesn't change that often. In such cases "flat" manifests or template-based generation can be simpler and faster.
Minimal Manifest Set
For a typical web service you essentially need three resources:
- Deployment — image, replicas, resources (limits/requests), readiness/liveness probes.
- Service — ClusterIP for internal traffic or LoadBalancer if the service must be externally accessible.
- Ingress (optional) — if you need HTTP/HTTPS routing, domain, SSL.
These can be generated from a template by substituting app name, image, and port. Many platforms, including Opsy AI, do exactly that: you describe the task ("deploy myapp from repo X, 2 replicas"), and the system outputs ready manifests. No Helm under the hood — just Deployment, Service, and Ingress when needed.
When Helm Is Still Needed
Helm makes sense when:
- you have many envs with different values (dev/staging/prod, different regions);
- there are shared dependencies — DBs, queues, common libs — and you want to version them with the app;
- you need rollbacks via
helm rollbackand release history; - the app is complex: lots of configs, init containers, dependencies between charts.
In other cases, describing the deploy as commands or short form ("deploy myapp with 2 replicas and 512Mi limit") with subsequent manifest generation can be faster and easier to maintain.
Practical Takeaway
Start with a minimal set. If the app is simple — don't drag Helm in "for growth". When multiple envs and complex dependencies appear, you can switch to Helm or Kustomize. Deploy in Kubernetes without Helm is possible and often appropriate; generating YAML from a description cuts time and reduces errors. More on DevOps and AI trends: DevOps and AI in 2026.