AI Gateway Deployment Models
AI Gateway offers two deployment models — SaaS and Hybrid — so you can match data residency requirements and infrastructure constraints.
| Where Can I Use This? | What Do I Need? |
|---|
- Prisma AIRS AI Gateway (Americas region)
|
- Prisma AIRS license with flex credits
- Strata Cloud Manager access
|
Prisma AIRS AI Gateway is available in two deployment models. In both models the management
plane lives in Strata Cloud Manager — you configure workspaces, LLM integrations, guardrails,
and routing policies in Strata Cloud Manager. The difference is where the data plane runs and therefore whether
request and response payloads pass through Palo Alto Networks infrastructure.
AI Gateway deployment model comparison
| SaaS | Hybrid |
|---|
| Data plane host | Palo Alto Networks (cloud-hosted) | Your Kubernetes cluster |
| Setup required | Enabled by default; no infrastructure setup | Kubernetes cluster, Helm chart installation |
| Request/response payloads | Routed through Palo Alto Networks infrastructure | Stay entirely within your environment |
| Management | Strata Cloud Manager | Strata Cloud Manager |
| Best for | Fast deployment; no data residency constraints | Regulated industries; strict data residency requirements |
SaaS deployment is the default model. When you activate AI Gateway, the SaaS data
plane is enabled immediately. Your developers send requests to the AI Gateway endpoint
(aigw.portkey.ai), and Palo Alto Networks infrastructure routes them to the appropriate
LLM provider. No additional infrastructure is required.
Hybrid deployment hosts the data plane in your own Kubernetes cluster. Request and
response payloads pass through your infrastructure and are never transmitted to Palo Alto Networks.
You download a values.yaml configuration file from Strata Cloud Manager
and deploy the gateway using Helm. The management plane, including guardrail policies, workspace
configuration, and observability dashboards, continues to run in Strata Cloud Manager.
You can run SaaS and Hybrid deployments simultaneously. For example, you might use the SaaS
model for general development workloads and a Hybrid deployment for production workloads that
handle sensitive data. Both deployments are managed from the same Strata Cloud Manager interface.
For configuration steps, see
Deploy AI Gateway.
