Services

ARK provides a rich ecosystem of services that extend the platform’s capabilities. These services are modular, Kubernetes-native components that can be installed and configured as needed.

Installing Services

Use make services to install and configure additional service capabilities. The command will scan for services with manifest files and let you choose to install, run in dev mode, or uninstall each service.

make services

Service Manifest Format

Each service must include a manifest.yaml file that describes the service and its capabilities:

name: ark-cluster-memory
description: In-memory storage with streaming support
supports:
  - dev
  - install
  - uninstall
post-install-instructions: |
  Service is now running at: http://ark-cluster-memory:8080

The post-install-instructions field is optional and will be displayed after successful installation.

Service Commands

Services follow a standard idiom using Make targets:

• make <service-name>-dev - Run the service in development mode locally
• make <service-name>-install - Build the container, push to cluster registry, and deploy using Helm
• make <service-name>-uninstall - Remove the service deployment from the cluster

The install command uses the scripts/build-and-push.sh helper to build and push container images to the cluster registry, then deploys the service using Helm charts.

Exposing Services

Services are exposed through Kubernetes Gateway API using HTTPRoute resources. The ARK quickstart sets up a gateway that services can use for external access.

Individual services can also be exposed using kubectl port-forward (as shown in post-install instructions).

Available Services

Core Services

• A2A Gateway - Agent-to-Agent Gateway implementing A2A protocol
• ARK Cluster Memory - In-memory storage with SSE streaming support
• Execution Engine - Framework adapter for Langchain and other AI frameworks

Execution Engines

• Langchain Execution Engine - Framework adapter for Langchain workflows

Development Tools

• ARK Dashboard - Next.js-based web interface for ARK management
• ARK API - REST API for managing ARK resources

Specialized Services

• Evaluator LLM - AI-powered query evaluation service using LLM-as-a-Judge
• Evaluation Operator - Automated evaluation and scoring of agent responses

Service Discovery and Annotations

ARK services use a standardized annotation system for discovery and integration:

Service Annotations

Helm Charts can be annotated to provide metadata for discovery:

# Chart.yaml
annotations:
  ark.mckinsey.com/service: "my-service"
  ark.mckinsey.com/resources: "service,deployment"

Annotation Schema