Agent-to-agent (A2A)

Documentation

Agentgateway

With agentgateway, you can route to agent-to-agent (A2A) servers and expose their tools securely.

Before you begin

Follow the Get started guide to install kgateway with agentgateway enabled.

Make sure that agentgateway is enabled in kgateway.

helm get values kgateway -n kgateway-system -o yaml

Example output:


agentgateway:
  enabled: true

Step 1: Deploy an A2A server

Deploy an A2A server that you want agentgateway to proxy traffic to.

Clone the kgateway repository.

git clone https://github.com/kgateway-dev/kgateway.git

From the root directory, build the sample A2A server.
```
VERSION=2.1.0-main make test-a2a-agent-docker
```
Load the image into your cluster. The following make command assumes that you are using a local Kind cluster.
```
CLUSTER_NAME=<your-cluster-name> VERSION=2.1.0-main make kind-load-test-a2a-agent
```

Deploy the A2A server. Notice that the Service uses the appProtocol: kgateway.dev/a2a setting. This way, kgateway configures the agentgateway proxy to use the A2A protocol.

kubectl apply -f- <<EOF
apiVersion: apps/v1
kind: Deployment
metadata:
  name: a2a-agent
  labels:
    app: a2a-agent
spec:
  selector:
    matchLabels:
      app: a2a-agent
  template:
    metadata:
      labels:
        app: a2a-agent
    spec:
      containers:
        - name: a2a-agent
          image: ghcr.io/kgateway-dev/test-a2a-agent:2.1.0-main
          ports:
            - containerPort: 9090
---
apiVersion: v1
kind: Service
metadata:
  name: a2a-agent
spec:
  selector:
    app: a2a-agent
  type: ClusterIP
  ports:
    - protocol: TCP
      port: 9090
      targetPort: 9090
      appProtocol: kgateway.dev/a2a
EOF

Step 2: Route with agentgateway

Route to the A2A server with agentgateway.

Create a Gateway resource that uses the agentgateway GatewayClass. Kgateway automatically creates an agentgateway proxy for you.

kubectl apply -f- <<EOF
apiVersion: gateway.networking.k8s.io/v1
kind: Gateway
metadata:
  name: agentgateway
spec:
  gatewayClassName: agentgateway
  listeners:
  - protocol: HTTP
    port: 8080
    name: http
EOF

Verify that the Gateway is created successfully. You can also review the external address that is assigned to the Gateway. Note that depending on your environment it might take a few minutes for the load balancer service to be assigned an external address. If you are using a local Kind cluster without a load balancer such as metallb, you might not have an external address.
```
kubectl get gateway agentgateway
```
Example output:
```
NAME           CLASS          ADDRESS                                  PROGRAMMED   AGE
agentgateway   agentgateway   1234567890.us-east-2.elb.amazonaws.com   True         93s
```

Create an HTTPRoute resource that routes incoming traffic to the A2A server.

kubectl apply -f- <<EOF
apiVersion: gateway.networking.k8s.io/v1
kind: HTTPRoute
metadata:
  name: a2a
spec:
  parentRefs:
    - name: agentgateway
      namespace: default
  rules:
    - backendRefs:
        - name: a2a-agent
          port: 9090
EOF

Step 3: Verify the connection

Get the agentgateway address.

export INGRESS_GW_ADDRESS=$(kubectl get gateway agentgateway -o=jsonpath="{.status.addresses[0].value}")
echo $INGRESS_GW_ADDRESS

kubectl port-forward deployment/agentgateway 8080:8080

Send a request to the A2A server.

curl -X POST http://$INGRESS_GW_ADDRESS:8080/ \
  -H "Content-Type: application/json" \
    -v \
    -d '{
  "jsonrpc": "2.0",
  "id": "1",
  "method": "tasks/send",
  "params": {
    "id": "1",
    "message": {
      "role": "user",
      "parts": [
        {
          "type": "text",
          "text": "hello gateway!"
        }
      ]
    }
  }
  }'

curl -X POST http://localhost:8080/ \
  -H "Content-Type: application/json" \
    -v \
    -d '{
  "jsonrpc": "2.0",
  "id": "1",
  "method": "tasks/send",
  "params": {
    "id": "1",
    "message": {
      "role": "user",
      "parts": [
        {
          "type": "text",
          "text": "hello gateway!"
        }
      ]
    }
  }
  }'

Example output:

{
  "jsonrpc": "2.0",
  "id": "1",
  "result": {
    "id": "1",
    "message": {
      "role": "assistant",
      "parts": [
        {
          "type": "text",
          "text": "hello gateway!"
        }
      ]
    }
  }
}

Cleanup

You can remove the resources that you created in this guide.

kubectl delete Deployment a2a-agent
kubectl delete Service a2a-agent
kubectl delete Gateway agentgateway
kubectl delete HTTPRoute a2a

AI for LLM consumption