Introducing Toolboxes in Foundry

Microsoft has launched the public preview of 'toolboxes' in Azure AI foundry, a feature that centralizes tool management for AI agents. Instead of each agent wiring its own tools with separate authentication and credentials, teams can define a named bundle of tools once in foundry, configure authentication centrally, and expose everything through a single MCP-compatible endpoint. Any agent runtime that supports MCP can consume the toolbox, including custom agents built with LangGraph or Microsoft Agent Framework, and coding agents like GitHub copilot. The public preview covers the 'build' and 'consume' pillars, with 'discover' and 'govern' features coming later. A step-by-step Python example shows how to create a toolbox with web search, Azure AI Search, and a GitHub MCP server, then attach it to an agent.