Skip to content
Prediction Guard is now available on the Azure Marketplace — Control & Govern AI from your VPC.
Get Started
Search icon
Get Started
 PRODUCT — Manage

Compose Sovereign AI Systems Without Compromising Security

Prediction Guard's Admin Console is your central control plane. Create and manage AI systems that consolidate models, MCP servers, tools, and agents. Self-hosted in your own infrastructure.

Book a DemoView Docs →
Self-Hosted
Control Plane
AWS · Azure · GCP
+ On-Prem + Air-Gap
Zero
Prompt Data Stored
 
 
 
Logo DimensionalPrediction Guard
Systems
Manage
Security
Analyze
Monitor
Govern
Audit
Settings
Users
Organizations
Systems
Manage your sovereign AI systems
Create System
Search systems...
 
 
PG System Staging
Status: Healthy
0
API Keys
0
Models
0
MCP Servers
Last update: Never
Created: 3/20/2026
Location: Staging
ID: ff2e9e54-13c4...
Manage
PG System V1
Status: Airgapped
0
API Keys
0
Models
0
MCP Servers
Last update: Never
Created: 3/20/2026
Location: kubernetes
ID: 96f015ce-bbbf...
Manage
PG System V2
Status: Never Connected
0
API Keys
0
Models
0
MCP Servers
Last update: Never
Created: 3/20/2026
Location: kubernetes
ID: afa137df-219e...
Manage
PG System V3
Status: Healthy
1
API Keys
6
Models
0
MCP Servers
Last update: 7 seconds ago
Created: 3/19/2026
Location: kubernetes
ID: 9c3f2a01-8d4b...
Manage
System Composition

Create a Sovereign AI System in Three Steps

An AI System consolidates your models, MCP servers, tool connections, and agents into a single governed unit — self-hosted in your own infrastructure.

  • Name your system and enable a public API endpoint with TLS hostname
  • Choose deployment: Kubernetes, Zero Dependency Binary, AWS, Azure, GCP, or Air-Gapped
  • The Admin Console generates a scoped kubectl command. Run it and Prediction Guard bootstraps in the predictionguard namespace
  • System appears as Healthy in the Admin Console, then add models, API keys, and governance policies
 
 
 
Create New AI System
Choose Quick Start for sensible defaults
Quick Start
Custom
System Name
production✓
Deployment Environment
Kubernetes
Selected
Binary
Air-Gap
Create System →
Model Management

Connect Any Model: Private, Managed, or External

Prediction Guard supports three model types, all managed from the same Admin Console without additional infrastructure. Connect any provider under one governed namespace.

  • External Models — Azure Foundry, AWS Bedrock, Google Vertex, OpenAI, Anthropic, and more
  • Managed Models — gpt-oss-120b, gemma-3-27b-it, Qwen2.5-Coder-14B hosted by Prediction Guard
  • Private Models — Deploy and connect models running on your own infrastructure
  • Each model inherits system-wide governance policies. No per-model policy configuration needed
 
 
 
All Models
External
Managed
Private
+ Add Model
Model Provider Type Status
gpt-oss-120b Prediction Guard Managed ● Active
gpt-4o OpenAI External ● Active
claude-3-sonnet Anthropic External ● Active
gemma-3-27b-it Prediction Guard Managed ● Active
One-Command Deployment

One kubectl Command:  Deployed in Your Infrastructure

The Admin Console generates a scoped installation command for your system. Run it on any machine with cluster access. Prediction Guard bootstraps in your own Kubernetes namespace, on-premise, or fully air-gapped.

  • Deploys to the predictionguard namespace automatically
  • System shows as Healthy once all services are running
  • Zero prompt data stored or logged by Prediction Guard at any point
  • Air-gapped deployments supported with offline packages
 
 
 
 
# Generated install command — scoped to your system
$ kubectl apply -f \
https://install.predictionguard.com/
?system=production \
&token=pg_sys_••••••••••••
Bootstrapping Prediction Guard services...
namespace/predictionguard created
deployment.apps/pg-control-plane created
deployment.apps/pg-gateway created
service/pg-api created
System production is Healthy — ready to use
Everything in Manage

Complete Control Plane for Your AI Infrastructure

 
Sovereign AI Systems
Compose systems that consolidate models, MCP servers, tools, and agents into one self-hosted unit.
 
Any Deployment
Kubernetes, Binary, AWS, Azure, GCP, or fully air-gapped. Your infrastructure, your choice.
 
15+ Model Providers
Connect Azure, AWS Bedrock, Vertex, OpenAI, Anthropic, and your own private models.
 
API Key Governance
Create and manage API keys with granular permissions per system, model, and tool.
 
AI Bill of Materials
Auto-generated, exportable BOM per system. Full inventory for compliance and governance.
 
Real-Time Monitoring
Live request volumes, latency, model performance, and resource utilization across all systems.

Ready to Own Your AI Stack?

See how Prediction Guard gives you full sovereignty over every model, agent, and API key in your organization.

Book a DemoAdmin Docs →