Microsoft Foundry Agent Service

Microsoft Foundry Agent Service is a fully managed enterprise AI agent platform in the Azure ecosystem, offering pay-per-use pricing for building, deploying, and orchestrating intelligent agents at scale. The platform addresses the growing need for sophisticated AI agent deployment in business environments by providing a comprehensive three-tier agent architecture that scales from no-code prototypes to fully custom containerized deployments. Prompt agents enable rapid prototyping through a portal-based no-code interface, workflow agents add declarative multi-step orchestration with branching and human-in-the-loop capabilities, and hosted agents allow teams to deploy custom code built with frameworks like LangGraph, Semantic Kernel, or AutoGen on managed infrastructure.

At the core of the platform is access to over 11,000 foundation models from the Foundry model catalog, including GPT-4o, Llama, Mistral, DeepSeek, and specialized industry models. Agents can intelligently route between models based on task requirements, and switching models is a configuration change rather than a code rewrite. Built-in tools cover web search via Bing, file search, Azure AI Search integration, code interpretation, and persistent memory, while enterprise connectors provide access to SharePoint, Microsoft Fabric, Azure Functions, and Logic Apps.

Enterprise security is deeply integrated through Microsoft Entra identity management with granular role-based access control, private virtual network isolation for compliance-sensitive workloads, and built-in content safety filters that include prompt injection and cross-prompt injection attack mitigation. The platform supports bring-your-own-resources for storage, search, and databases, enabling organizations to maintain data sovereignty and governance.

Multi-agent orchestration is a core differentiator, supporting patterns such as sequential handoff, group chat, delegation, and parallel execution through AutoGen and Semantic Kernel integration. Workflow agents add visual declarative branching and conditional logic, with human-in-the-loop approval steps available inside the same agent graph. End-to-end observability is provided through OpenTelemetry-based tracing surfaced in Application Insights, with dashboards for performance, cost, and safety metrics that make debugging and audit review practical at production scale.

The platform is best suited for enterprise development teams already invested in the Microsoft and Azure ecosystem, organizations requiring strict compliance and governance for AI deployments, and companies looking to consolidate agent development under a single managed service with deep integrations into Microsoft 365, Teams, and Azure DevOps.