Getting Started

Why the Enterprise Kit?

The only enterprise governance framework spanning all layers of the Microsoft AI Agent Platform Stack:

Azure AI Foundry provides the runtime for AI agents. Agent 365 provides the governance control plane. But enterprises need more — cryptographic audit trails, automated certification pipelines, MCP fleet lifecycle management, cross-agent policy topology, and enterprise system adapters before any agent touches production. This kit enhances Agent 365 with these enterprise-grade controls that Microsoft does not provide natively.

Without Enterprise Kit	With Enterprise Kit
6–12 month build cycle	2–4 week deployment
Custom adapter code per system	7 pre-built MCP adapters + hosting templates
No audit trail	Hash-chained Decision Ledger (NIST/EU AI Act)
Manual compliance checks	Automated policy enforcement + Agent 365 governance
Ad-hoc evaluation	Continuous eval with CI/CD integration
Bespoke infra scripts	28 Bicep modules, subscription-scoped (CAF-aligned)
No MCP governance	Full MCP fleet lifecycle: versioning, auth, hosting, catalog
No certification gates	Bronze → Silver → Gold automated promotion pipeline

Assess Your Readiness

Not sure where to start? Take the 5-minute Enterprise AI Governance Assessment.

Evaluate your organization across 8 governance dimensions — identity & access, observability, compliance, tool governance, evaluation, business value, data boundaries, and agent lifecycle. Get a personalized maturity score (L1–L4) with specific recommendations for which kit capabilities to adopt first.

Take the Assessment →

Overview

The Azure Foundry Enterprise Kit is a production-grade accelerator for deploying governed AI agents on the full Microsoft AI Agent Platform Stack — Azure AI Foundry Agent Service, Microsoft Agent Framework, Agent 365 governance, Foundry Toolbox & MCP, and the Three-IQ intelligence layer. It provides a complete stack — from 28 Bicep infrastructure modules to trust extensions, MCP fleet governance, and continuous evaluation.

Prerequisites

Azure subscription with Contributor + User Access Administrator role
Azure CLI v2.60+ installed and authenticated (az login)
Bicep CLI v0.30+ (az bicep version)
Python 3.11+ (for SDK starters, adapters, and azure-ai-evaluation)
Node.js 20+ (for TypeScript SDK starter and @azure/ai-projects)
Resource providers registered: Microsoft.MachineLearningServices, Microsoft.CognitiveServices, Microsoft.Search, Microsoft.DocumentDB, Microsoft.ApiManagement
Azure AI Foundry Hub provisioned (or let the Bicep template create one)

Quick Start

1. Clone the Repository

git clone https://github.com/your-org/azure-foundry-enterprise-kit.git
cd azure-foundry-enterprise-kit

2. Install Python Dependencies

pip install -e ".[adapters,dev]"
# Includes: azure-ai-projects, azure-ai-evaluation,
# azure-identity, azure-monitor-opentelemetry

3. Deploy Azure Infrastructure (28 Modules)

Provisions Azure resources across 28 Bicep modules — Foundry Hub, AI Search, Cosmos DB, Key Vault, App Insights, Content Safety, APIM AI Gateway, API Center, MCP hosting, VNet, Managed Identities, and more:

az deployment sub create \
  --location eastus2 \
  --template-file infra/bicep/main.bicep \
  --parameters infra/bicep/parameters/dev.bicepparam

4. Configure Agent Manifest

Copy an agent template and customize the manifest for your use case:

cp -r agent-templates/knowledge-agent my-agent
# Edit my-agent/agent-manifest.yaml

5. Configure MCP Toolbox

# Select auth method: key, agent-identity, managed-identity, oauth, unauthenticated
python sdk-starters/python/toolbox_manager.py \
  --config toolbox-configs/knowledge-agent-toolbox.yaml \
  --auth-mode managed-identity

6. Run Evaluations

Execute evaluations via the Azure AI Foundry Evaluations SDK (azure-ai-evaluation):

python sdk-starters/python/eval_runner.py \
  --manifest my-agent/agent-manifest.yaml \
  --eval-pack eval-packs/builtin/builtin-evaluators.yaml

7. Publish Agent

Publish the agent as a Foundry Agent Application and register it in Agent 365:

# Deploy as ARM-native Agent Application
python trust-extensions/agent_application_publisher.py \
  --manifest my-agent/agent-manifest.yaml

# Register in Agent 365 for cross-platform governance
python trust-extensions/agent365_publisher.py \
  --manifest my-agent/agent-manifest.yaml

Project Structure

Directory	Purpose
`adapters/`	7 MCP adapters: ServiceNow, SAP, Salesforce, Workday, MS Graph, Internal API, MCP hosting
`agent-templates/`	5 pre-built agent blueprints deployed as Foundry Agent Applications
`infra/bicep/`	28 Bicep modules (CAF-aligned): Foundry, Search, Cosmos DB, VNet, APIM, Content Safety, API Center, MCP hosting, cost management, hub-spoke, etc.
`eval-packs/`	Built-in + custom evaluators via azure-ai-evaluation SDK
`policy-packs/`	5 declarative policy packs: identity, toolbox, data boundary, memory retention, escalation
`trust-extensions/`	9 modules: Decision Ledger, Scorecard, Cert Pipeline, Policy Graph, Agent365 Publisher, Toolbox Lifecycle Manager, etc.
`sdk-starters/`	Python (azure-ai-projects 2.0) + TypeScript (@azure/ai-projects) SDK clients + MCP auth provider
`observability/`	Azure Monitor Workbooks, 6 KQL query packs (incl. MCP fleet dashboards), telemetry schema
`toolbox-configs/`	Foundry Toolbox allowlist configs per agent class
`zero-trust/`	17 Zero Trust controls: identity, network, data, runtime, MCP security, governance, compliance, residency

Next Steps

Take the Maturity Assessment — Understand your governance gaps and get a prioritized adoption plan
Read the Architecture Guide — Microsoft AI Agent Platform Stack alignment, 22 modules, MCP Fleet Governance
Review Agent Blueprints — 5 pre-built templates with manifest-driven configuration
Follow the Deployment Guide — Full Azure provisioning with environment profiles
Set up Trust & Governance — Decision Ledger, Certification Pipeline, Agent 365 integration
Explore MCP Fleet Governance — Toolbox lifecycle, auth provider, hosting templates
Review Zero Trust — 15 security controls from identity to MCP security