Skip to main content

Containers for GenAI & Agentic AI

The Open-Source Way

Serve, package, and ship local LLMs and agentic systems — on any OCI runtime, no paid Docker Desktop.

Container-native, not Docker-native

One compose.yaml runs identically on Colima, OrbStack, Rancher Desktop, or Podman. Built on the OCI standard and the Compose Spec — free of paid Docker Desktop.

One real use case, built step by step

Go from a served model to naive RAG → Agentic RAG → a multi-agent crew, hand-authoring the stack service by service. Every module adds one deliberate step.

Runs on a 16 GB laptop

Small quantized models and the native-server / containerized-app pattern keep every lab to ~4–6 GB. No cloud GPU required; Apple Silicon M1 or Windows 11 + WSL2 both work.