Microsoft Ignite 2025: Microsoft Foundry - The Unified Enterprise AI Platform
Microsoft Ignite 2025 unveiled Microsoft Foundry - a unified enterprise platform integrating agents, models, tools, and governance for building AI at scale. Explore Foundry Agent Service, Foundry Tools, Foundry IQ, and the complete AI development stack.
Overview
I just returned from Microsoft Ignite in San Francisco, and if you’re building AI agents for enterprise, this was the event you needed to attend.
Microsoft unveiled Microsoft Foundry - not just a collection of services, but a unified platform that brings together agents, models, tools, and governance under one roof. This is Microsoft’s vision for enterprise AI operations at scale.
This article breaks down Microsoft Foundry’s core components and what they mean for enterprise AI development.
Key Insight: Microsoft Foundry is to enterprise AI what Azure was to cloud computing - a complete infrastructure layer that makes enterprise operations possible.
Table of Contents
- What is Microsoft Foundry?
- Foundry Agent Service
- Foundry Tools
- Foundry IQ
- Foundry Models
- Foundry Control Plane
- Microsoft Agent Framework
- Bonus: Azure API Management as AI Gateway
- Why This Matters
- Resources
What is Microsoft Foundry?
Microsoft Foundry is a unified Azure platform-as-a-service offering for enterprise AI operations, model builders, and application development. It combines production-grade infrastructure with developer-friendly interfaces, enabling teams to focus on building applications rather than managing infrastructure.
Core Components
Microsoft Foundry consists of below integrated components:
- Foundry Models - Access to 1,900+ AI models
- Foundry Agent Service - Build, host, and scale intelligent agents
- Foundry Tools - 1,000+ curated tools with MCP support
- Foundry IQ - Unified knowledge layer powered by Azure AI Search
- Foundry Control Plane - Centralized governance and observability
Key Characteristics
- Unified Management: Single management grouping with built-in enterprise-readiness
- Integrated Capabilities: Tracing, monitoring, evaluations, and customizable configurations
- Streamlined Operations: Unified RBAC, networking, and policies under one Azure resource provider
Resources:
Foundry Agent Service
Foundry Agent Service is the centerpiece for building, deploying, and scaling AI agents securely within the Microsoft Foundry ecosystem.
Key Features
Foundry Agent Service enables you to build, deploy, and scale AI agents with built-in memory, tools, and enterprise governance
Design Multi-Agent Workflows with Fine-Grained Control: Quickly design and deploy multi-agent applications with a user-friendly interface and comprehensive toolset.
Develop agents with your choice of language models, frameworks, and protocols—and bring your custom-code agents into Microsoft Foundry with ease
Build with leading frameworks and models, and securely host your custom agents: Develop agents with your choice of language models, frameworks, and protocols—and bring your custom-code agents into Microsoft Foundry with ease.
Secure and govern agents with Microsoft Entra Agent ID, centralized control plane observability, continuous guardrails, and compliance-ready security for enterprise-scale operations
Meet Security and governance requirements: Secure and govern agents with Microsoft Entra Agent ID, centralized control plane observability, continuous guardrails, and compliance-ready security for enterprise-scale operations.
Foundry Agent Service Capabilities
1. Hosted Agents
Deploy custom-code agents into fully managed runtime:
Supported Frameworks:
- Microsoft Agent Framework
- LangGraph
Resources:
2. Built-in Memory
Enterprise-grade agent memory system:
Capabilities:
- Retains chat history, preferences, and task outcomes
- Secured with Entra ID
- User-scoped memory instances
- Multi-agent collaboration without data crosstalk
- Custom topics for optimized memory retrieval
Resources:
3. Multi-Agent Workflows
Visual orchestration of complex, multi-step processes:
Design Options:
- Visual designer in Foundry portal
- VS Code extension
- Code-first with Microsoft Agent Framework
Features:
- Specialized agents collaborate
- State sharing across agents
- Error recovery
- Long-running operation support
- Connected agents with intelligent delegation
Resources:
4. One-Click Publishing to Microsoft 365
Deploy directly to where users work:
Deployment Targets:
- Microsoft 365 Copilot
- Teams Chat
- Multiple M365 channels from single build
Streamlined Configuration:
- Entra ID integration
- Azure Bot Service setup
- Admin Center routing
- Central IT approval
Resources:
Foundry Tools
A unified hub for discovering, testing, and governing the tools that make agents intelligent and actionable.
What is Foundry Tools?
Foundry Tools provides access to 1,000+ curated Microsoft and partner tools through a centralized catalog with support for open standards.
Supported Standards
- MCP (Model Context Protocol) - Connect to MCP servers
- A2A (Agent-to-Agent) - Enable agent-to-agent communication
- OpenAPI - Integrate REST APIs seamlessly
Key Capabilities
Public Tool Catalog:
- 1,000+ curated Microsoft and partner tools
- Real-time data access (Bing, SharePoint, Azure AI Search)
- Business actions (Logic Apps, Azure Functions, OpenAPI)
- Productivity apps integration
Private Organizational Catalogs:
- Secure internal API management
- Compliant reuse across teams
- Custom tool registration
Flexible Authentication:
- Microsoft Entra Agent Identity
- OAuth passthrough
- Managed Identity
Governance Integration:
- Azure Policy integration via AI Gateway
- Runtime guardrails for tool calls
- Complete observability
Resources:
- Foundry Tools Overview
- MCP Documentation
- MCP Registry
- Microsoft Learn MCP Server GitHub
- Azure AI Foundry MCP Server
Foundry IQ
Foundry IQ, powered by Azure AI Search, is Microsoft Foundry’s centralized knowledge layer that connects agents to enterprise data.
What is Foundry IQ?
Instead of wiring retrieval logic into every agent, you define reusable knowledge bases around topics. Any number of agents and applications can then connect and be grounded with that same knowledge base.
Architecture
Federated Data Sources:
- Microsoft 365 SharePoint
- Fabric IQ
- OneLake (Azure Data Lake)
- Azure Blob Storage
- Azure AI Search indexes
- Web (Bing)
- MCP servers (private preview)
Agentic Retrieval Engine
A self-reflective query engine that uses AI to plan, search, and synthesize answers:
Query Execution Flow:
- Source Selection - Determines which sources to query
- Query Planning - Decomposes into subqueries
- Federation - Executes against selected sources
- Ranking - Semantic ranker scores and filters results
- Reflective Search - Reviews results and issues follow-up queries
Performance:
- 36% improvement in response relevance vs. traditional RAG
- Configurable “retrieval reasoning effort”
- Multi-hop query planning
Automatic Indexing Pipeline
For indexed sources, Foundry IQ handles:
- Content ingestion
- Chunking
- Vectorization
- Hybrid retrieval preparation
- Layout-aware enrichment (with Content Understanding)
Security & Compliance
- Automatic Purview enforcement on every query
- Built-in security vs. fragmented systems
- User access permissions maintained
- Enterprise-ready compliance
Resources:
Foundry Models
Access to 1,900+ AI models through a comprehensive catalog organized by capability and provider.
Model Categories
Models Sold Directly by Azure:
- GPT-4o, GPT-4, GPT-3.5 (Azure OpenAI)
- Claude family (Anthropic) - NEW at Ignite 2025
- High integration with Azure services
- Microsoft Responsible AI standards
- Enterprise-grade scalability
Partner & Community Models:
- Anthropic Claude (NEW at Ignite 2025):
- Claude 3.5 Sonnet - Ideal balance of intelligence, speed, and cost
- Claude 3.5 Haiku - Fast and cost-effective for high-volume use cases
- Claude 3 Opus - Most capable for complex reasoning tasks
- Extended context window up to 200K tokens
- Computer use and multimodal capabilities
- Unified Azure billing and enterprise security
- Meta (Llama family)
- xAI
- DeepSeek
- Hugging Face models
- Specialized and domain-specific models
Model Router (Generally Available)
Dynamically select the best-fit model for each prompt:
- Balances cost, performance, and quality
- Available in Foundry Agent Service (public preview)
- Particularly helpful for multi-agent systems
Fine-Tuning Support
Agents can use fine-tuned models directly:
- Improves reasoning and task accuracy
- Synthetic data generation from APIs
- Teaches models enterprise context
Resources:
Foundry Control Plane
Foundry Control Plane brings identity, controls, observability, and security together so teams can build, operate, and govern agents with confidence.
Foundry Control Plane provides unified governance, controls, and observability for all your AI agents
Unified Governance
Fleet-Wide Visibility:
- See all agents (Foundry-native, Microsoft, and third-party)
- Models and tools in one place
- Health, cost, performance metrics
- Risk and policy coverage
Agent Identity:
- Microsoft Entra Agent ID for every agent
- Durable identity and lineage
- Access governance
Controls & Guardrails
Comprehensive Protection:
- Input controls - Validate incoming requests
- Output controls - Check agent responses
- Tool call controls - Govern tool invocations
- Tool response controls - Validate tool outputs
Built-in Guardrails:
- Task adherence
- Sensitive data detection
- Groundedness checks
- Prompt injection mitigation
- Protected materials detection
Observability
Integrated Monitoring:
- Built-in evaluations
- OpenTelemetry-based tracing
- Continuous red teaming
- Azure Monitor integration
Framework Support:
- Microsoft Agent Framework
- LangGraph
- OpenAI Agents SDK
AI Red Teaming Agent (Generally Available):
- Continuously test agents against real-world risks
- Automated vulnerability detection
- Security posture assessment
Security Integration
Deep Integration:
- Microsoft Defender - Real-time risk detection
- Microsoft Purview - Data protection
- Microsoft Entra Agent ID - Specialized identity for AI agents
Upcoming Integrations:
- Palo Alto Networks - Runtime protection
Agent Identity Concepts
Microsoft Entra Agent ID is a specialized identity type designed specifically for AI agents, providing:
Core Capabilities:
- Standardized framework for governing, authenticating, and authorizing AI agents
- Secure access to resources across Microsoft services
- User interaction management with proper authorization
- System-to-system communication security
Integration Benefits:
- Automatic agent identity creation and management
- Lifecycle management from development to production
- Simplified permission management
- Enhanced security and auditability
- Seamless integration with Microsoft Foundry
Microsoft Foundry automatically integrates with Microsoft Entra Agent ID, creating and managing agent identities throughout their entire lifecycle, ensuring secure operations while maintaining governance standards.
Resources:
Bonus: Azure API Management as AI Gateway
Microsoft Agent Framework
The foundation of Foundry Agent Service is the Microsoft Agent Framework - the unification of Semantic Kernel and AutoGen into one open-source SDK:
Key Features:
- Durable execution - Agents persist and recover automatically
- Built-in security - Native Microsoft Purview SDK integration
- UI compatibility - Supports AG-UI and ChatKit standards
- Multi-language - Full support for .NET and Python
- Cross-platform - Local and cloud deployment
- Enterprise features - Thread-based state management, type safety, middleware
Core Capabilities:
- Individual agents with LLM processing
- Tool and MCP server integration
- Graph-based multi-agent orchestration
- Type-based routing and nesting
- Human-in-the-loop support
Resources:
Azure API Management as AI Gateway
One of the most important but often overlooked announcements was Azure API Management as the AI Gateway for Microsoft Foundry.
What is AI Gateway?
Azure API Management serves as a centralized gateway that secures and manages all AI application traffic, including:
- Model API calls
- Agent interactions
- Tool invocations (including MCP tools)
_Azure API Management as AI Gateway provides centralized security, governance, and observability for all AI
Key Capabilities
1. Security & Authentication
- Centralized authentication for all AI services
- Azure Entra ID integration
- API key management
- OAuth token handling
2. Governance & Compliance
- Azure Policy integration - Enforce organizational policies
- Content filtering and moderation
- Usage quotas and rate limiting
- Audit logging for compliance
3. Observability
- Centralized monitoring across all AI services
- Request/response logging
- Performance metrics
- Cost tracking and attribution
4. MCP Tool Management
- Register and catalog MCP tools
- Apply policies to tool invocations
- Monitor tool usage patterns
- Version management
5. Azure API Center Integration
- MCP Registry - Centralized catalog of tools
- Discovery across the organization
- Consistent governance policies
- Usage analytics
Why This Matters
Before AI Gateway:
- Each agent manages its own security
- Policies scattered across systems
- No centralized monitoring
- Tool governance fragmented
With AI Gateway:
- Single point of control
- Consistent policy enforcement
- Unified observability
- Enterprise-grade security
Resources:
- Lab: Governing AI Apps & Agents with AI Gateway
- AI Gateway Documentation
- MCP Register in Azure API Center
Resources
Microsoft Ignite 2025 - Complete Labs & Skill Plans
**Microsoft Ignite 2025: Next Steps - All Labs & Skill Plans **
Your comprehensive hub for all Microsoft Ignite 2025 hands-on labs, skill plans, and learning paths across Azure AI, Microsoft Foundry, Security, and more.
Core Documentation
Microsoft Foundry:
Components:
Agent Framework:
GitHub Repositories
Core Projects:
MCP Resources:
Samples & Labs:
Key Ignite Sessions
Foundation:
- Innovation Session: Build & Manage AI Apps with Your Agent Factory
- AI agents in Azure AI Foundry, ship fast, scale fearlessly
Deep Dives:
- AI powered automation & multi-agent orchestration
- AI builder’s guide to agent development in Foundry Agent Service
- AI tools for agents: Plug, play and reach
- The future of RAG with agentic knowledge retrieval and AI Search
Blog Posts
Platform Announcements:
Technical Deep Dives:
- Foundry Agent Service at Ignite 2025
- Foundry Control Plane
- Foundry IQ: Unlocking Knowledge
- Multi-Agent Orchestration
Getting Started
For Beginners:
For Developers:
For Enterprise:
Conclusion
Microsoft Foundry is the complete infrastructure layer for enterprise AI—unifying agents, models, tools, knowledge, and governance under one platform. With 1,900+ models, any framework support, built-in security, and one-click Microsoft 365 deployment, it transforms AI from experiments to production-ready operations. The future of enterprise AI is here.
Have questions or want to share your experience with Microsoft Foundry? Connect with me on LinkedIn or drop a comment below.