A2AServer

A2AServer enables hosting external agent frameworks within ARK using the Agent-to-Agent (A2A) protocol.

Specification


apiVersion: ark.mckinsey.com/v1prealpha1
kind: A2AServer
metadata:
  name: aws-operator-agent
spec:
  # Service address for the A2A server.
  # Supports value, valueFrom.serviceRef, valueFrom.configMapKeyRef, valueFrom.secretKeyRef
  address:
    value: http://ark-agentcore-bridge.default.svc.cluster.local:80/a2a/agent/aws_operator_agent-jg0yD9Hv2n
  # Human-readable description of the A2A server
  description: AWS operations agent with read-only access to AWS services
  # How often to poll the server for updates (default: 1m)
  pollInterval: 1m
status:
  conditions:
    # Ready: A2AServer is reachable and operational
    - type: Ready
      status: "True"
      reason: AgentDiscovered
      message: Successfully discovered agent
    # Discovering: Agent discovery is in progress
    - type: Discovering
      status: "False"
      reason: DiscoveryComplete
      message: Agent discovery completed
  # Last successfully resolved server address
  lastResolvedAddress: http://ark-agentcore-bridge.default.svc.cluster.local:80/a2a/agent/aws_operator_agent-jg0yD9Hv2n

Examples

Created Agent Example

When the A2AServer above is created, the controller automatically creates an Agent:


apiVersion: ark.mckinsey.com/v1alpha1
kind: Agent
metadata:
  name: aws-operator-agent
  annotations:
    # A2A server that created this agent
    ark.mckinsey.com/a2a-server-name: aws-operator-agent
    # Resolved address of the A2A server
    ark.mckinsey.com/a2a-server-address: http://ark-agentcore-bridge.default.svc.cluster.local:80/a2a/agent/aws_operator_agent-jg0yD9Hv2n
    # Skills discovered from the A2A server
    ark.mckinsey.com/a2a-server-skills: '[{"name":"describe_ec2_instances","description":"List and describe EC2 instances in the account"}]'
spec:
  description: AWS operations agent with read-only access to AWS services
  prompt: You are aws_operator_agent. AWS operations agent with read-only access to AWS services
  executionEngine:
    name: a2a
  # No modelRef - A2A agents don't require models

Behavior

When an A2AServer is created:

Discovery: Controller connects to the server and discovers available agents. Tries /.well-known/agent-card.json (A2A v0.3+), then /.well-known/agent.json (A2A v0.2.x).
Agent Creation: For each discovered agent, an Agent resource is created with:
- Owner reference to the A2AServer
- executionEngine.name: a2a
- Annotations identifying the A2AServer
Status Updates: Controller continuously monitors server health