OK AI — AI Workloads & Agentic Operations
Your Kubernetes Framework. Built for the Age of AI.
OK AI is the AI capability area of OpenKubes: run self-hosted AI services on Kubernetes and explore explicitly constrained, evidence-driven agent integrations. It is a platform capability, not a separate Kubernetes distribution or a separate ok-ai source repository.
OK AI Chat — first UI preview
Explore the OK AI Chat UI prototype → — an interactive visual preview with sample responses only. It does not send any questions to an AI service or public API. No login required.
OK AI roadmap
Explore our three-phase OK AI roadmap: Documentation Assistant → Read-only Diagnostics → Agentic Operations. The Documentation Assistant has passed an isolated private lab format-and-source-ID integration check; the public service is not yet released.
Explore OK AI
| Capability | Purpose | Documentation |
|---|---|---|
| Open WebUI | Self-hosted chat frontend for AI backends | Open WebUI |
| OpenClaw | Optional agent backend through the Agent Interface Contract | OpenClaw |
| kagent & Platform Diagnostics | Read-only diagnostics provider behind the HTTP/OpenAPI contract | kagent & Platform Diagnostics |
| Agentic Operations | Proposal, policy, authorization, execution and evidence boundaries | Agentic Operations |
| AI Glossary | Beginner-friendly explanations of AI concepts, from LLMs to agents, RAG and evals | AI Glossary |
Architecture at a glance
OpenKubes separates the management plane, AI workloads, model-serving endpoints and optional agents. The framework defines contracts; selected ecosystem components implement them.
- Management plane: Crossplane claims can reconcile AI services onto authorized workload clusters.
- AI workload cluster (
ok-ai): Hosts AI-facing services such as Open WebUI where provisioned. - Model backend: GPU-backed services such as Ollama can be provided separately, subject to infrastructure and network configuration.
- Agent boundary: Optional agents consume explicitly defined interfaces and must not silently inherit unrestricted Kubernetes access.
The repository's AI platform overview documents the architecture and historical deployment examples. Treat examples as environment-specific until prerequisites and versions have been verified.
Start here
- New to OpenKubes? Use Getting Started to choose Local or Cluster first.
- Have a Kubernetes workload cluster? Read Open WebUI and its upstream component README.
- Evaluating agents? Read OpenClaw, kagent & Platform Diagnostics, Agentic Operations, and the relevant ADRs before live tests.
Source ownership: AI components are maintained in openkubes/openkubes/platform/ai/; cluster-specific deployment inputs belong to the respective operator environment and repository. This website is a guided entry point, not a replacement for the source-owned runbooks.