What is Microsoft Foundry Agent Service?
Microsoft Foundry Agent Service is a managed Microsoft Azure platform for building, deploying, and scaling AI agents. An agent combines a model from the Foundry model catalog with instructions and tools, and unlike a simple chatbot, it can independently complete multi-step tasks, call external systems, and make decisions.
The central interface is the Responses API, through which any framework or your own code gets access to Foundry models and platform tools. The service supports two agent types: Prompt agents, which are defined entirely through configuration in the Foundry portal, via SDK, or REST, and are run completely by Foundry (no application code, no compute to manage), and Hosted agents, where you write your own agent code — for example with Microsoft Agent Framework, LangGraph, the OpenAI Agents SDK, or the Anthropic Agent SDK — package it as a container, and let Foundry run it with a managed endpoint, autoscaling, and a dedicated Microsoft Entra identity.
Core Features
- Responses API as a unified entry point for model access and tool orchestration.
- Built-in tools: Code Interpreter, File Search, Web Search, Memory, plus integration of external MCP servers (Model Context Protocol) and custom functions.
- Two operating models: fully managed Prompt agents and container-based Hosted agents running your own code.
- Observability: end-to-end tracing of model calls and tool usage, with Application Insights integration.
- Enterprise security: Microsoft Entra agent identities, role-based access control, content filters/guardrails against prompt injection, and network isolation via virtual networks.
- Publishing: agent versioning and distribution through Microsoft Teams, Microsoft 365 Copilot, and the Entra Agent Registry, plus agent-to-agent communication via the A2A protocol (preview).
Typical Use Cases
Customer service agents can check orders, initiate returns, and interact with CRM, ERP, or ticketing systems to resolve requests without manual intervention.
Internal assistants for employees answer questions about company policies, help coordinate schedules, or create reports by executing multi-step workflows through connected tools.
Data analysis agents accept natural language questions, use Code Interpreter for calculations, and File Search or connected knowledge sources for context, delivering insights to business teams without deep technical expertise.
Agents intended for existing Microsoft 365 environments can be deployed directly to Teams or Microsoft 365 Copilot through the platform’s publishing features.
Benefits
- Two operating models for different needs: fast start with Prompt agents, full control with Hosted agents.
- Unified access to many models from the Foundry model catalog without switching integration layers.
- Broad tool ecosystem, including open standards like MCP.
- Enterprise-grade identity, security, and networking capabilities built in from the start.
- End-to-end observability for debugging, evaluation, and operating agents in production.
Integration with innFactory
As a Microsoft Solutions Partner, innFactory supports you with Microsoft Foundry Agent Service: agent design, tool and MCP integration, security architecture, and production deployment as either a Prompt or Hosted agent.
Contact us for a non-binding consultation on Foundry Agent Service and Microsoft Azure.
Typical Use Cases
Frequently Asked Questions
What is Microsoft Foundry Agent Service?
Foundry Agent Service is a managed Microsoft Azure platform for building, running, and scaling AI agents. It supports fully managed Prompt agents that require no application code, as well as Hosted agents where you run your own agent code in a Foundry-managed container.
When should I use Foundry Agent Service?
The service is a good fit when an LLM needs to independently solve multi-step tasks, call tools, or reach out to external systems, rather than just generate text. For simple, conversational bots with fixed dialog paths, a classic bot service is often a better fit.
What does Foundry Agent Service cost?
There are no separate platform fees for creating and running Prompt agents; you are billed for model token consumption and the tools you use. Hosted agents are additionally billed based on compute consumption (vCPU and memory hours).
Which tools can agents use?
Agents can access built-in tools such as Code Interpreter, File Search, Web Search, and Memory, integrate custom functions via function calling, and connect to external MCP servers (Model Context Protocol). Available tools and their preview status can vary by region and model.
How do Prompt agents differ from Hosted agents?
Prompt agents are defined through configuration (portal, SDK, or REST) and run entirely by Foundry, with no compute to manage. Hosted agents run as your own container with your agent code, for example built with Agent Framework, LangGraph, or the OpenAI Agents SDK, and are hosted, scaled, and given a dedicated Microsoft Entra identity by Foundry.
Can Foundry Agent Service be used in a GDPR-compliant way?
When using European Azure regions with appropriate data residency configuration, the service can be operated in a GDPR-compliant manner. Microsoft offers Data Processing Agreements for this; the exact setup depends on the chosen architecture and the tools you connect.
Note: All product information on this page has been compiled with care, but is provided without guarantee and may be outdated or incomplete. Cloud services evolve rapidly — features, pricing, SLAs, and availability change frequently. Authoritative and up-to-date information can only be found on the official product page of Azure (official documentation). This page does not represent an offer by Azure.
