A data clean room
for software.

from agentlockbox import AgentLockbox, GoogleCloud, OpenSourceModel

client = AgentLockbox()

# The manifest points to your existing agent image
deployment = client.deploy_kubernetes(
    name="client-production",
    manifest="./k8s/claims-agent.yaml",
    model=OpenSourceModel(repository="Qwen/Qwen3-8B"),
    target=GoogleCloud(
        project="client-prod",
        region="us-central1",
    ), # Uses your configured cloud identity
)

print(deployment.url)

Choose the model; we deploy it with your software in the confidential cluster.

from agentlockbox import AgentLockbox, GoogleCloud

client = AgentLockbox()

# A regular service works too; no model is required
deployment = client.deploy_docker(
    name="private-service",
    dockerfile="./service/Dockerfile",
    context="./service",
    target=GoogleCloud(
        project="client-prod",
        region="us-central1",
    ),
)

print(deployment.url)

The SDK builds and deploys a standard containerized application.

# Add AgentLockbox to your existing Terraform stack
resource "agentlockbox_kubernetes_deployment" "claims" {
  name          = "client-production"
  manifest_path = "./k8s/claims-agent.yaml"
  model_repository = "Qwen/Qwen3-8B"

  google_cloud = {
    project = "client-prod"
    region  = "us-central1"
  }
}

output "agent_url" {
  value = agentlockbox_kubernetes_deployment.claims.url
}

A Terraform provider could manage the cluster, workload, and selected model as part of your infrastructure stack.