Kubestack catalog deprecationThe registry at kbst.xyz is deprecated and will be shut down on December 31, 2026. Migration guide

KubestackKubestack
v1.0.0-beta.1

It's February 2027
you built the Kubernetes platform
application teams love to use

Now stop daydreaming and

How it works
platform-repo · agent session
youAdd a GPU node pool to the EKS apps cluster
agentedited   eks_zero_node_pool_gpu.tf  +18
agentran      tofu fmt && tofu validate ✓
agentopened   PR #142 "Add GPU node pool"
[Plan: 3 to add, 0 to change, 0 to destroy.]
reviewapproved by a platform engineer ✓
pipelineapplied to ops ✓
pipelinepromoting to apps after merge ●

GitOps workflow

Agents edit. Humans review. The pipeline applies.

Agents only edit and validate local files, the GitOps pipeline owns all changes to real infrastructure, keeping non-deterministic LLM behavior away from production.

How the GitOps process works

  1. 01 · Agent

    Edit & validate

    Local files only

  2. 02 · Reviewer

    Review the plan

    Pull request with tofu plan

  3. 03 · Pipeline · now

    Apply to ops

    Validates for real after merge

  4. 04 · Pipeline

    Promote to apps

    On apps-deploy tag

  5. 05 · Pipeline

    Promote to apps-prod

    On apps-prod-deploy tag

it learns the skill, then scaffolds your platform

How it works

short instructions for you and your agent

Why Kubestack?

Building a Kubernetes based platform requires tight integration of countless components across multiple environments. Maintaining such a platform without infrastructure as code is tedious and error prone.

Coding agents change how you write that infrastructure as code, but they do not change what makes it reliable. An agent working on raw OpenTofu quickly produces a slightly different approach in every session. An agent that has learned the Kubestack skill works inside a proven framework instead: one canonical way to scaffold, configure and evolve your platform, and a GitOps pipeline that owns every change to real infrastructure.

Kubestack is the AI native way to have your agent define your entire stack in one OpenTofu code base and continuously evolve your platform safely using GitOps.

Learn how Kubestack can help you manage the complexities of building and managing Kubernetes platforms

Platform teams using Kubestack

logo_woven-by-toyota
logo_techminers
logo_pento
logo_luminovo

Proudly sponsored by

Accession1

Self-service provisioning of resources and access for people and AI agents

accession.one ↗ (opens in a new tab)

One code base

Clusters, node pools and platform features in OpenTofu

One canonical way to do every task, so agent sessions produce consistent, reviewable code instead of a different approach every time
eks_zero_cluster.tf
module "eks_zero" {
  source = "github.com/kbst/terraform-kubestack//aws/cluster"
 
  configuration {
    apps {
      name_prefix      = "kbst"
      cluster_min_size = 3
    }
    ops {}
  }
}

Inheritance

Teams and platforms of any size

In platform engineering, there is no one size fits all. No matter if single engineer, or entire platform team. Using Kubestack your agent safely and reliably controls every aspect of your platform architecture. From cloud providers to regions, from environments to clusters, and from node pools to platform features.
opsappsapps-prod

Review

GitOps workflow and pipeline

Agents are great at changing code, but you should not let them apply infrastructure changes. Kubestack strictly separates the two: your agent edits and validates files locally, the GitOps pipeline reviews, validates and applies every change. This keeps non-deterministic LLM behavior from ever reaching your running platform directly.

kubestack / plan · ops

+ 3 to add · ~ 0 · − 0

module.eks_zero_node_pool_gpu

Platform features

Platform features and guides

The real value of platforms lies in making powerful, seamlessly integrated features available to application developers. The Kubestack skill teaches your agent to integrate modules for popular Kubernetes services in production ready configurations. And platform engineering guides help you deliver common platform features end to end.
nginxprometheuscert_manager

For your agent

One skill teaches it the framework

AI native by design: your coding agent learns the Kubestack skill and works within the framework's conventions instead of improvising raw OpenTofu
kubestack.com/SKILL.md

The best Kubernetes for each cloudbuild unified platforms using EKS, AKS and GKE

Kubestack builds on each cloud's managed Kubernetes offering for the best possible integration and support, but unifies platform features, developer experience and runtime environments for your application developers across all clusters and clouds. Your agent uses the same conventions everywhere, so a multi-cloud platform is no harder to maintain than a single-cloud one.

Unified across all clusters and clouds

Platform features · developer experience · runtime environments

Managed Kubernetes

Amazon EKS

AWS

Managed Kubernetes

Azure AKS

Microsoft Azure

Managed Kubernetes

Google GKE

Google Cloud
Kubestack· same conventions
Each cloud’s own managed Kubernetes underneath, one unified platform on top, one set of conventions for your agent.

it learns the skill, then scaffolds your platform

How it works

short instructions for you and your agent