Workshops: Agentic Coding by Tech Stack
Kubernetes & Platform Engineering with AI Agents: Workshop
Platform teams maintain the Helm charts, Kustomize overlays, CRDs, operators and golden paths that every product team relies on. This advanced workshop shows how Claude Code, OpenAI Codex, Cursor or GitHub Copilot accelerate that work while GitOps, policies and local test clusters keep the agent far from cluster-admin rights. It goes much deeper than our general DevOps and IaC workshop.
Who is this workshop for?
- Platform engineering and internal developer platform teams
- Kubernetes operators and cluster administrators
- Go developers writing controllers and operators
- Teams running GitOps with Argo CD or Flux
- SREs responsible for multi-cluster operations and upgrades
What you will take away
- Use GitOps as a natural guardrail: agents change desired state through pull requests and the GitOps controller applies it.
- Have agents author and refactor Helm charts and Kustomize overlays, verified by linting, rendering and schema validation.
- Build a small operator with controller-runtime and kubebuilder with agent support, including envtest-based tests.
- Write admission and security policies with Kyverno or OPA Gatekeeper and let agents test them against real manifests.
- Prepare Kubernetes upgrades by having agents find deprecated APIs and draft migration pull requests.
Agenda
Module 1
Safe setup
- Local clusters with kind or k3d as the agent's sandbox
- Read-only kubeconfigs, RBAC for agent identities, no production contexts
- Cluster access via MCP: what to allow and what never to allow
Module 2
Charts and manifests
- Exercise: turning a grown Helm chart into a reusable library chart
- Kustomize overlays per environment without drift
- Validation with helm lint, rendered templates and schema checks
Module 3
GitOps flow
- Argo CD or Flux: agent pull requests against the config repository
- Promotion between environments and rollback
- Reviewing rendered diffs rather than template diffs
Module 4
Operators and CRDs
- Designing CRDs and reconcile loops with the agent
- Typical agent mistakes: missing idempotency, wrong status handling
Module 5
Policies and upgrades
- Policy as code with Kyverno or Gatekeeper, tested by the agent
- Finding deprecated APIs before a cluster upgrade
- Golden path templates, for example for Backstage
Why GitOps and agents fit together
In a GitOps setup nobody runs kubectl apply against production by hand, and an agent should not either. It works in the configuration repository, opens a pull request and exposes the rendered manifests; Argo CD or Flux applies the change only after review. That gives you an audit trail, easy rollback and a clear boundary for the agent.
Where agents need a short leash
Kubernetes YAML looks simple, which makes plausible but wrong output dangerous: a missing resource limit, a privileged container or an overly broad RBAC role slips through a quick review easily. We combine policies, rendering and review so that such errors fail automatically.
- No cluster-admin, no write access to production clusters
- Secrets via External Secrets or Sealed Secrets, never in plain text
- Limits, probes and security contexts enforced by policy
- Logs and events treated as untrusted input because of prompt injection
Built for certified environments
jocapps, the company behind neooutput, is ISO 27001 certified, and our AI Solutions Architects design agent workflows that fit your change management. We work in your configuration repositories or a sandbox, remotely or on-site.
Frequently asked questions
How does this differ from your DevOps and IaC workshop?
The DevOps workshop covers Terraform, CI/CD and operations broadly. This one assumes Kubernetes experience and goes deep into Helm, GitOps, operators, policies and cluster upgrades.
Should agents have kubectl access at all?
On local and test clusters, yes, so they can verify their own work. On shared or production clusters we recommend read-only access for diagnosis at most, with all changes flowing through GitOps.
Does this work with OpenShift, Rancher or managed Kubernetes?
Yes. The patterns apply to EKS, AKS, GKE, OpenShift, Rancher and European managed Kubernetes offerings. We adapt the exercises to your distribution and GitOps tooling.
Workshop
Get in touch
Should your platform team ship faster with agents, without any agent touching production?
Tell us roughly how many people will join and pick dates that work – we will get back to you within 24 hours with a proposal for content, format and schedule.
Prefer to talk first?
15 minutes, free and without obligation – with our AI Trainers and AI Code Engineers.
Book a call for free (opens in new tab) +49 761 519 751 50 (24/7)Request a training / workshop
Workshop: Kubernetes & Platform Engineering with AI Agents: Workshop · 1 day
More workshops
- Workshops: Agentic Coding by Tech StackRuby on Rails with AI Coding Agents: Hands-On Workshop
- Workshops: Agentic Coding by Tech StackElectron & Tauri Desktop Apps with AI Coding Agents: Workshop
- Workshops: Agentic Coding by Tech StackAgentic Coding with React & Next.js: Workshop for Dev Teams
- Workshops: Agentic Coding by Tech StackAgentic Coding for Angular Enterprise Apps: Workshop
- Workshops: Agentic Coding by Tech StackAgentic Coding with Vue & Nuxt: Hands-On Workshop
- Workshops: EU AI Act & ComplianceShadow AI: How to Detect and Govern It – Workshop
