deployment live host cluster do-nyc1-v12-nyc2 served

V12 · sample page

Agent Boilerplate
is up and serving.

A reusable platform for building LLM agents — not one agent. Standing up a new industry or customer means writing a manifest, a connector and a prompt pack. It should never mean touching code under packages/.

The three bets

01

Hexagonal architecture

@v12/core declares ports and imports no infrastructure. Every external system is an adapter chosen at boot, and a dependency-cruiser rule fails the build if that ever stops being true.

02

Agent behaviour is data

A versioned manifest compiles into a LangGraph graph. Models are referenced by alias, never by id — retargeting a vertical onto a different model is a row update.

03

Types are the enforcement

Zero any, branded ids, exhaustive unions, Result for expected failures. The frontend imports the same Zod schemas the API validates with.

This deployment

hostboilerplate.workspace.v12.dev
clusterdo-nyc1-v12-nyc2 · DigitalOcean Kubernetes · nyc1
namespaceboilerplate
workload2 × nginx-unprivileged, static content from a ConfigMap
exposureIngress → ClusterIP :80 → container :8080
purposePlaceholder — confirms DNS, ingress and TLS before the real app ships