✨ The Containment Era is here. Secure AI workloads before they breach. →The Containment Era is here. →The Containment Era is here. →Explore ✨
What is a VCA?
From catalog to contained — in under four hours.
A Validated Containment Architecture (VCA) is a lab-tested deployment blueprint that contains the AI platform you already run. No design work. No app changes. No application-team sign-off. Three steps:
Lab-validated · Free 30-day Enterprise trial · Provision in 15 minutes · Deploy your first VCA this afternoon
Find your AI platform in the catalog — Bedrock, Azure AI Foundry, Gemini, GitHub, MCP.
Run the Terraform straight from GitHub. Your workloads never change — and never notice.
Policies start in monitor mode. Enforcement is one flag flip — rollback is the same flag.
Blast radius: one governed workload. Enforced before any breach is ever detected.
Without containment architecture, every AI workload
is an ungoverned blast radius.
Without Containment
With Containment Architecture
Pick the path that matches where you are.
Every Validated Containment Architecture deploys on Aviatrix Enterprise — including the Controller, CoPilot, and Distributed Cloud Firewall. Start with a free trial or deploy directly from GitHub if you already run Enterprise today.
Enterprise is free — every VCA included on day one
Subscribe to Aviatrix Enterprise on AWS or Azure Marketplace, finish provisioning in under 15 minutes, and deploy any live architecture within the trial window.
30-day free trial · No contract · Full platform included
Deploy any live VCA from GitHub
Click into any live architecture below to find the blueprint for your environment — Terraform, Helm, or policy YAML. Each detail page has the full deployment guide.
Blueprint available on every live VCA · No additional license required
Validated Containment Architectures
Contain AWS Bedrock AgentCore
Zero-trust egress for Bedrock AgentCore agents at the network layer. AI-aware SmartGroups map Bedrock workloads dynamically; default-deny MCP egress means each agent declares its allowed destinations as policy-as-code.
Contain Azure AI Foundry Agents
Containment for Azure AI Foundry deployments. AI-aware SmartGroups target ai_agent resource types directly; east-west segmentation prevents agent-to-agent lateral movement and every flow is logged for compliance.
Contain Enterprise MCP (Obot)
Policy-as-code containment for Obot-based MCP infrastructure. FirewallPolicy CRDs let each MCP server declare its network scope; a compromised server cannot exfiltrate through ungoverned paths.
Contain Enterprise GitHub Pipelines
URL-path scoping for CI/CD and coding-agent workloads. Differentiate github.com/acme-corp/* from github.com/*. Blocks runtime package installs from npm, PyPI, and Docker Hub.
Contain Enterprise AI Chat
Production containment for enterprise AI chat. Managed WebGroups govern which LLM providers are reachable; AI workload URL categories block file-sharing and unauthorized providers by default.
Containment Plugin for Microsoft Agent Control Specification
Compile your .guardrails.yaml into Aviatrix DCF rules via the open-source acs-to-dcf shim. No Terraform blueprint — policy plugin, not a deployable architecture.
Contain Google Gemini Enterprise Agent Platform
Network containment for Gemini Enterprise Agent Platform agents (formerly Vertex AI). PSC-I insertion, default-deny egress, named shadow-model deny, and east-west isolation — one DCF policy model across managed runtime and GKE shapes.
Contain AI Agent Harnesses — OpenClaw / NemoClaw
Default-deny egress for OpenClaw, NemoClaw, and Hermes agent runtimes on AWS. Agent-class presets map coding, research, support, and regulated-data agents to separate DCF policy packs. Shadow-model deny fires at priority 10.
Start the trial.
Deploy your first VCA this afternoon.
Subscribe to Aviatrix Enterprise on AWS or Azure Marketplace, finish provisioning in under 15 minutes, then deploy any Wave 1 architecture from GitHub.
Controller 8.1+ · 1 managed network per deployment · No developer compliance required