Accelerating Enterprise AI with Microsoft Foundry

Today at Ignite 2025, the spotlight shone brightly on Microsoft Foundry the unified AI platform empowering enterprises to build, deploy, and govern intelligent agents at scale. New features released today are designed to enhance model accessibility, agent sophistication, and enterprise-grade reliability. Below is a deep dive into these powerful updates.

Anthropic’s Claude Models Join Foundry

Microsoft Foundry customers can now access three Claude models: Claude Sonnet 4.5, Claude Opus 4.1, and Claude Haiku 4.5. These frontier AI models bring specialized capabilities to Foundry users, including advanced code generation, complex reasoning with ultra-long context (up to 1 million tokens), and robust tool-use skills. Critically, the addition of Claude means Azure becomes the only cloud platform to offer customers both OpenAI’s GPT series and Anthropic’s Claude models side-by-side, giving enterprises unprecedented flexibility to choose the optimal model for each use case.

Foundry Models: Unparalleled Choice and Performance

Ignite revealed major enhancements in the Foundry Models suite, empowering organizations with:

• Broader model catalog: Over 11,000 models now available including frontier generators and specialized AI from Anthropic, Cohere, NVIDIA, RosettaFold, MongoDB’s Voyage AI, and more.

Microsoft Foundry Model Catalog

• Model Router (GA): Automatically selects the best model for each prompt based on real-time checks for latency, quality, and cost. Benchmarks show up to 50% latency reduction, saving time and money. 
• Priority Processing (Public Preview): SLA-backed, low-latency inference lanes for critical workloads using GPT‑4.1, with promising additions like GPT‑5 on the horizon. 
• Fine-Tuning: Preview tools that allow custom tuning using synthetic data, adapter layers, and community-hosted models—ranging from Ministral 3B to Llama 3.3—enable tailored enterprise intelligence. 
• Synthetic Data Generation (Preview): Generate Synthetic Data within Microsoft Foundry for fine-tuning by providing sample data. More information on this available here Generate synthetic data with Microsoft Foundry (Preview) - Microsoft Foundry | Microsoft Learn 

• Knowledge: Provides a single endpoint that connects agents to multiple knowledge sources Azure Blob, OneLake, SharePoint, Azure AI Search indexes, the web (via Bing), and more without needing separate RAG pipelines for each agent.

Microsoft Foundry Knowledge Base

Next‑Gen Agent Hosting with Foundry Agent Service

Foundry Agent Service received significant enhancements that elevate intelligent agent development and deployment:

• Hosted Agents (Public Preview): Deploy agents built with Microsoft Agent Framework, LangGraph, or other third-party tools directly into a fully managed runtime no container orchestration needed.
• Multi-Agent Workflows (Public Preview): Visually design and program multi-step, collaborative agent workflows. Ideal for complex tasks like onboarding or finance approvals, with built-in fault recovery and orchestration. 

Workflow Builder in Microsoft Foundry with Sequential, Human in the loop, and Group Chat pattern samples available for quick start development. You can also start with a Blank Workflow

• Built-in Memory (Public Preview): Agents now persist context securely across sessions. Developers define memory scope, while Access is governed via Entra ID, ensuring session continuity without compromising data isolation. 
• Foundry Tools (Public Preview): A catalog of 1,000+ prebuilt and private connectors (MCP, OpenAPI) and even and Browser Automation for UI based tasks. Agents can securely call business systems like SAP, UiPath, and much more.

Tools Catalog for that empowers Agents to connect to tools such as Azure SQL MCP Server, Hugging Face, Github, and many more

OpenAPI, MCP, and A2A Integrations available in Foundry as Custom tools

• Foundry IQ (Public Preview): Integrates Azure AI Search to provide agents with RAG-powered access to enterprise sources (SharePoint, ADLS, web), enforcing Purview security policies automatically. 
• Agent Evaluation: enables developers to benchmark and validate agent performance across tasks using automated metrics, custom test sets, and real-world scenarios ensuring reliability and continuous improvement.

Agent Evaluations with Pre-built templates with the ability to also create custom Evaluations for your own Agents to measure performance and quality

Enterprise Confidence with Foundry Control Plane

Security, compliance, and governance the hallmarks of any enterprise-grade platform are front and center in today’s announcements:

• Foundry Control Plane (Public Preview): Centralizes identity (Entra Agent ID), policy enforcement, observability, and risk control. Includes safeguards for inputs, outputs, tool calls, and prompt hygiene. 

Foundry Control Plane providing a unified view for all operations for all resources within the Foundry Project including model and agent activities

• Deeper integrations: Defender and Purview enrich agent telemetry. GitHub Advanced Security + Defender for Cloud bring dev-to-production safety. 
• AI Gateway: an essential part of Microsoft Foundry for enforcing enterprise grade governance, usage control, and observability across model and agent interactions.
• One-click Agent 365 publication: Simplified deployment path from Foundry Agent Service into Microsoft 365 Copilot and Teams managed intelligently via Agent 365 in the Admin Center. 

Enterprises gain a unified control plane to manage AI agents across development, production, and deployment ensuring trust, visibility, and fault-free governance.

Hybrid AI: Local, Cloud & Edge for Maximum Reach

Now with Microsoft Foundry’s, Teams can deploy AI experiences everywhere from cloud to edge to on-device ensuring performance, privacy, and resilience regardless of connectivity.

• Foundry Local expanded: Now in private preview on Android, joining Windows and macOS support.
• Managed Instance on Azure App Service (Public Preview): Enables .NET apps to migrate easily, tapping into Foundry’s model, tool, and governance stack. 

Looking Ahead

Today’s Ignite announcements make Microsoft Foundry a robust platform for enterprise intelligence bringing model diversity, rich agent orchestration, strong governance, and hybrid deployment flexibility.

Enterprises now have a mature foundation to:

• Experiment across best-in-class AI models,
• Seamlessly embed agents into business processes,
• Maintain robust control through centralized governance, and
• Empower users regardless of device or deployment environment.

Foundry’s full stack from models to control plane becomes a key enabler of scalable, secure, and intelligent transformation. With these layered improvements from model innovation to multi-agent workflows, from on-device edge support to unified governance Foundry is well-positioned to lead enterprises into a new era of purposeful AI.

Get Started Today

Microsoft Foundry is available today in Azure. Below are helpful sessions from Ignite to showcasing all the possibilities and capabilities of this Platform. 

Innovation Session: Build & Manage AI Apps with Your Agent Factory

What is Microsoft Foundry? - Microsoft Foundry | Microsoft Learn

AI powered automation & multi-agent orchestration in Microsoft Foundry

Innovation Session: Build & Manage AI Apps with Your Agent Factory