Skip to content

Deploying AGT on Google Cloud GKE¶

No Azure Required — AGT is pure Python with zero cloud-vendor dependencies. It runs anywhere containers run: GCP, AWS, on-prem, or your laptop.

Prerequisites¶

gcloud CLI configured with GKE permissions
Docker installed locally
A GKE cluster running

1. Dockerfile¶

FROM python:3.12-slim
WORKDIR /app
COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt
COPY . .
EXPOSE 8080
CMD ["python", "-m", "agent_os.server", "--port", "8080"]

2. Build & Push to Artifact Registry¶

gcloud auth configure-docker us-central1-docker.pkg.dev

docker build -t agt-governance .
docker tag agt-governance:latest \
  us-central1-docker.pkg.dev/<PROJECT>/agt/agt-governance:latest
docker push us-central1-docker.pkg.dev/<PROJECT>/agt/agt-governance:latest

3. GKE Deployment¶

apiVersion: apps/v1
kind: Deployment
metadata:
  name: agt-governance
spec:
  replicas: 2
  selector:
    matchLabels:
      app: agt-governance
  template:
    metadata:
      labels:
        app: agt-governance
    spec:
      containers:
        - name: agt
          image: us-central1-docker.pkg.dev/<PROJECT>/agt/agt-governance:latest
          ports:
            - containerPort: 8080
          env:
            - name: AGT_POLICY_PATH
              value: /app/policies/
            - name: AGT_LOG_LEVEL
              value: audit
          resources:
            requests:
              cpu: 250m
              memory: 512Mi
            limits:
              cpu: 500m
              memory: 1Gi
---
apiVersion: v1
kind: Service
metadata:
  name: agt-governance
spec:
  type: ClusterIP
  selector:
    app: agt-governance
  ports:
    - port: 80
      targetPort: 8080

4. Deploy¶

kubectl apply -f agt-deployment.yaml
kubectl rollout status deployment/agt-governance

Using with Google ADK Agents¶

AGT includes a native Google ADK adapter. Wrap your ADK agent's tool calls with PolicyEvaluator.evaluate() — see examples/quickstart/google_adk_governed.py.

See Also¶