Skip to main content

Meet the Kubernetes Agent

Welcome to k8s-autopilot, TalkOps's intelligent way to handle your entire Kubernetes lifecycle. Instead of writing endless YAML or fighting with Helm templates, you can interact with a dedicated team of AI agents built specifically for cluster operations.

We designed k8s-autopilot to feel less like a rigid script and more like a senior DevOps colleague. Whether you need to generate a complex Helm chart from scratch, securely deploy to a live cluster, or onboard a new microservice into ArgoCD, the agent handles the heavy lifting while keeping you firmly in control.

Why we built this​

Managing Kubernetes at scale is tough. Junior engineers often hit a steep learning curve, while senior architects find themselves drowning in repetitive runbooks, troubleshooting minor YAML indentation errors, or fighting with out-of-sync cluster states.

We wanted to fix this by combining the reasoning power of Large Language Models (LLMs) with the strict reliability of tools you already trust.

With k8s-autopilot, you get:

  • Instant chart generation: Describe your application, and the agent writes production-ready Helm templates.
  • Active cluster awareness: The agent doesn't guess; it uses the Model Context Protocol (MCP) to read live cluster states and existing releases before making decisions.
  • Human-in-the-loop safety: No black-box operations. Before any command touches your cluster or ArgoCD, the agent pauses, shows you a "dry run" plan, and waits for your explicit approval.
  • Self-healing: If a generation fails validation, the agent catches it, reads the error log, and fixes its own YAML dynamically.

How it works under the hood​

The underlying architecture relies on a Supervisor-Worker pattern powered by LangGraph. Think of the Supervisor as the project manager, fielding your requests and delegating them to highly specialized experts.

When you ask the system to "Deploy the checkout API to ArgoCD," the Supervisor routes the task to the ArgoCD Orchestrator. The orchestrator checks if the project and repository exist, generates the necessary sync manifests, and presents it to you. Only when you click approve does it execute the changes.

Key capabilities at a glance​

  1. Helm Chart Generation: Give it a natural language prompt, and it outputs a beautifully structured Helm chart with values.yaml and README documentation.
  2. Active Helm Management: You can install, upgrade, rollback, and search Helm releases on live clusters directly through conversation.
  3. ArgoCD Onboarding: Effortlessly connect new repositories, spin up projects, and sync applications. The agent can even pull logs to troubleshoot failing ArgoCD deployments.

Getting Started​

The easiest way to take k8s-autopilot for a spin is via Docker.

# Pull the latest image
docker pull sandeep2014/k8s-autopilot:latest

# Run the agent (requires your OpenAI key and local kubeconfig)
docker run -d -p 10102:10102 \
  -e OPENAI_API_KEY=your_key \
  -v ~/.kube/config:/root/.kube/config \
  --name k8s-autopilot \
  sandeep2014/k8s-autopilot:latest

The agent is entirely model-agnostic. You can use OpenAI (gpt-4o), Anthropic (claude-3.5-sonnet), or Google (gemini-1.5-pro) just by tweaking your .env file configuration.

Got complex requirements? You can even route different sub-agents to different models (e.g., using a faster model for the Supervisor and a heavy reasoning model for the Code Generator).

What's next?​

Take a look at the rest of the documentation for deep dives into Helm generation and ArgoCD workflows. If you want to contribute or track our progress, come hang out with us on GitHub.