KubeCopilot Documentation¶
Welcome to the official documentation for KubeCopilot — the pluggable, engine-agnostic AI agent platform for Kubernetes and OpenShift.
-
:material-rocket-launch:{ .lg .middle } Getting Started
Install the operator, deploy an agent, and start chatting in minutes.
-
:material-cog:{ .lg .middle } Configuration
Custom skills, AGENT.md instructions, BYOK providers, and runtime settings.
-
:material-shield-account:{ .lg .middle } Multi-Tenant Sessions
Namespace-per-tenant isolation with NetworkPolicy, RBAC, and finalizer cleanup.
-
:material-file-tree:{ .lg .middle } Architecture
CRD-driven design, request flow, and pluggable engine architecture.
-
:material-api:{ .lg .middle } Agent Server
API contract, webhook payloads, background tasks, and how to create new engines.
-
:material-github:{ .lg .middle } Contributing
Development setup, project structure, and contribution guidelines.
What is KubeCopilot?¶
KubeCopilot is a Kubernetes operator that deploys and manages AI agents inside your cluster, controlled entirely through CRDs. Unlike read-only AI assistants, KubeCopilot agents can reason, plan, and execute — running kubectl commands, managing resources, and automating complex operations autonomously.
The operator is engine-agnostic: swap the AI backend by changing a container image in your CR — no code changes, no redeployment. Skills, instructions, custom agents, and even the LLM model can be reconfigured at runtime through the Web UI, without restarting a single pod.
KubeCopilot runs on both vanilla Kubernetes and OpenShift, with a native OpenShift Console Plugin that embeds the chat UI directly into the web console.
Experimental
This project is experimental and has not been tested in a production environment. Running AI agents with cluster access carries inherent risks. Review all manifests, RBAC rules, and agent instructions carefully before deploying.
Screenshots¶
