Microsoft Foundry Agent Service is a managed service for building, deploying and scaling AI agents without managing the underlying compute and storage. Microsoft’s AI Agents for Beginners course presents it as the newer of two Microsoft routes to implementing the Function calling, and positions it for enterprise applications on the grounds that it is fully managed and offers enterprise-grade security.
The course states three advantages over calling a model API directly: tool calling is automatic, so the application does not parse a tool call, invoke the tool and handle the response itself, all of which happens server-side; conversation state is managed through threads rather than by the developer; and a set of tools is available out of the box for reaching common data sources.
Capabilities
Tools are divided into two categories. Knowledge tools retrieve information — grounding with Bing Search, File Search, and Azure AI Search. Action tools do something — function calling, Code Interpreter, tools defined from an OpenAPI specification, and Azure Functions.
Tools are combined into a toolset, which may mix functions the developer supplies with prebuilt ones; the model decides which to use for a given request. The course’s worked example builds a toolset from a custom SQLite query function and the Code Interpreter tool, then creates an agent with a model, name, instructions and that toolset, so the model chooses between querying the data and computing over it depending on what is asked. Threads hold the message history of a particular conversation.
Adoption & Ecosystem
The course illustrates the service through a sales-data scenario, in which a conversational agent answers questions about an organization’s sales figures by querying them and analysing the results. It appears alongside Microsoft Agent Framework as the managed alternative to running an agent framework yourself.