Built for the Cluster

A multi-service orchestration platform routing LLM inferences through secure execution pods, connected via resilient NATS JetStream messaging.

Brain

The core LLM orchestrator. Holds conversations, reasons about tasks, and decides what to delegate.

Multi-turn conversations with Claude
Up to 10 tool-use iterations per response
Multi-model support and role-based routing

Isolation by Default

LLM applications are migrating from single-process bots to multi-service platforms. Production demands isolation, audit trails, and uptime guarantees that no generic chatbot framework provides.

Task Pods: Run strictly separated container processes for dangerous workflows.
Workers: Handle long-running inference without blocking chat IO.
Security Posture: Non-root execution, seccomp RuntimeDefault, and read-only filesystems.

metadata:
  name: "task-pod-isolation"
spec:
  containers:
  - name: "executor"
    securityContext:
      runAsUser: 1000
      runAsNonRoot: true
      readOnlyRootFilesystem: true
      allowPrivilegeEscalation: false