# Microsoft Foundry Microsoft Foundry is an enterprise AI platform for building, grounding, deploying, and governing AI applications and agents. It helps development and data teams connect models, organizational knowledge, business tools, and lifecycle controls in Azure. - Findings24: - Site web: - Tarifs: Freemium - Catégories: Infrastructure pour agents IA, Plateformes d’inférence IA, Déploiement de modèles d’IA ## Qu'est-ce que Microsoft Foundry ? Microsoft Foundry is an enterprise AI platform for building, grounding, deploying, and governing AI applications and agents. It combines model access and customization, agent development, organizational knowledge, business tools, observability, and trust controls in a unified Azure environment. The platform is designed for development and data teams creating AI-driven, cloud-native applications. Teams can choose from a broad model catalog, build agents with Foundry Agent Service or the open-source Microsoft Agent Framework, connect agents to organizational and business-system context, and manage production behavior through monitoring and governance capabilities. Foundry pricing is consumption-based across the services and features used. Each service has its own billing model and price, so the total cost depends on the selected capabilities and usage. ## Que peut faire Microsoft Foundry ? ### Broad model catalog and model customization Access foundational, open, reasoning, multimodal, and industry-specific models from providers listed by Microsoft, then train, fine-tune, distill, or upgrade models with minimal coding. ### Real-time model routing Use intelligent model routing to select a model for a particular task and optimize performance or cost without rewriting the application. ### Agent development and hosting Build action-oriented, context-aware agents with Foundry Agent Service and use the open-source Microsoft Agent Framework to create, orchestrate, and extend agent applications. Foundry also supports serverless containers and event-driven functions for hosting and scaling. ### Grounding and business-system connectivity Ground agents in organizational intelligence with permission-aware grounding and Microsoft Graph integration. Connect workflows to business systems through more than 1,400 prebuilt connections, including SAP, Salesforce, and Dynamics 365, or add custom tools through Model Context Protocol. ### Observability and evaluation Trace agent interactions in production with OpenTelemetry-based tooling and evaluate applications using built-in measures for coherence, relevance, groundedness, and safety. ### Unified trust and governance controls Manage agents, tools, and knowledge sources through a control plane integrated with Microsoft Entra, Purview, and Defender. Runtime controls can help simulate, detect, and mitigate prompt attacks, hallucinations, and sensitive-data leakage. ## Cas d'utilisation ### Automate complex business processes Create agents that perform specific, action-oriented tasks across connected business systems while retaining human control over the workflow. ### Build grounded enterprise assistants Combine permission-aware organizational context, Microsoft Graph, and business data so applications can reason using relevant organizational signals rather than relying only on a general model. ### Develop and modernize AI applications Use model choice, routing, agent frameworks, prebuilt tools, and Azure services to add AI capabilities to new or existing cloud-native applications. ### Operate AI agents in production Trace interactions, evaluate response quality and safety, and apply runtime trust controls as agent applications move from development into ongoing use. ### Create data and document workflows Combine customizable tools such as OCR, translation, speech, and object detection with agents or applications to automate information-processing workflows. ## Questions fréquentes ### What is Microsoft Foundry used for? It is used to build, ground, deploy, monitor, and govern AI applications and agents. Common workflows include business-process automation, data and document processing, application modernization, and adding AI capabilities to cloud-native applications. ### Can Microsoft Foundry use models from different providers? Yes. Microsoft describes Foundry Models as a catalog spanning foundational, open, reasoning, multimodal, and industry-specific models, including models from providers such as OpenAI, Anthropic, Meta, Google, xAI, and Hugging Face. ### How can agents access business context and tools? Agents can use permission-aware grounding, Microsoft Graph integration, Microsoft IQ, Foundry IQ, and Fabric IQ. They can also connect to business systems through prebuilt connections, use customizable tools, or access custom APIs through Model Context Protocol. ### How does Microsoft Foundry support production oversight? Its control-plane capabilities include OpenTelemetry-based tracing, evaluators for coherence, relevance, groundedness, and safety, and runtime controls intended to help detect and mitigate prompt attacks, hallucinations, and sensitive-data leakage. ### How is Microsoft Foundry priced? Pricing is consumption-based and depends on the individual Foundry services and features used. Each service has its own billing model and price; Microsoft provides pricing estimates, a pricing calculator, and sales support for quotes. ## Informations rapides - **Category**: Enterprise AI platform - **Platform**: Microsoft Azure - **Primary users**: Developers, data scientists, and organizations building AI applications and agents - **Development options**: Foundry Agent Service, Microsoft Agent Framework, prebuilt tools, and custom MCP tools - **Governance**: Unified control plane with Microsoft Entra, Purview, and Defender integrations - **Pricing model**: Consumption-based; individual services and features have separate billing models