CAIRA provides several baseline configurations for Azure AI Foundry based solutions, so users can have a consistent, scalable, reliable deployment of Azure AI Foundry and supporting infrastructure in an accelerated time frame in support of agentic workloads.

This configuration is designed as a simple development environment for common AI workloads such as generative AI application development, building autonomous AI agents capable of performing complex tasks, as well as the evaluation and testing of AI models.

Deploys Azure AI Foundry with basic setup

• Project and deployment for getting started
• Public networking
• Microsoft-managed file storage
• Microsoft-managed resources for storing Agents threads and messages

See the reference architecture at reference_architectures/foundry_basic.

This provides the same features of basic but exposes the Foundry APIs via private endpoints.

Deploys Azure AI Foundry with basic private setup

• Project and deployment for getting started
• Bring-your-own private networking
• Microsoft-managed file storage
• Microsoft-managed resources for storing Agents threads and messages

See the reference architecture at reference_architectures/foundry_basic_private.

This configuration includes everything in the Basic setup and adds explicit capability host connections for data services so you can control where agent data is stored and how it's accessed.

Deploys Azure AI Foundry with explicit host connections for data sovereignty and compliance

• Project and default model deployments
• Explicit agent capability host connections to Azure Cosmos DB, Azure AI Search, and Azure Storage
• Bring-your-own or module-provisioned dependent services
• Identity-first defaults (RBAC), ready for enterprise hardening
• Built-in observability via Log Analytics and Application Insights

See the reference architecture at reference_architectures/foundry_standard.

This configuration includes everything in the standard setup but exposes the Foundry APIs as well as the capability host resources via private endpoints.

Deploys Azure AI Foundry with explicit host connections for data sovereignty and compliance

• Project and default model deployments
• Explicit agent capability host connections to Azure Cosmos DB, Azure AI Search, and Azure Storage
• Bring-your-own or module-provisioned dependent services
• Bring-your-own network for capability hosts and agents to be connected to.
• Identity-first defaults (RBAC), ready for enterprise hardening
• Built-in observability via Log Analytics and Application Insights

See the reference architecture at reference_architectures/foundry_standard_private.

• foundry_basic - Public networking with portal deployment support
• foundry_basic_private - Basic configuration with network isolation
• foundry_standard - Enterprise features with Bring Your Own Resource (BYOR) approach
• foundry_standard_private - Enterprise configuration with protected agent support

Foundry-only use case: If you only need Azure AI Foundry (account, project, model deployments, observability) and do NOT plan to use the Agent Service, choose one of the _basic reference architectures (foundry_basic for public endpoints or foundry_basic_private for private networking). The _standard variants add agent-focused dependent resources (Cosmos DB, Storage, AI Search) and capability host connection wiring that are not required when not using the Agent service.

For detailed networking requirements, security configurations, and deployment prerequisites, refer to the official Azure AI Foundry documentation.

For Infrastructure-as-Code implementation details, see the individual reference architecture folders in this repository.