Implement generative AI and agentic solutions 2.2 Build agents by using Foundry
Twelve agents, built as Foundry prompt agents, LangGraph apps, and Microsoft Agent…
Study guide skill: Integrate agent tools, including APIs, knowledge stores, search, content understanding, and custom functions
Question B2-009
Twelve agents, built as Foundry prompt agents, LangGraph apps, and Microsoft Agent Framework apps, all need the same MCP servers, OpenAPI tools, and Azure AI Search. The platform team wants to manage credentials and guardrails for these tools in one place and change tools without modifying or redeploying the agents. What should you use?
Select one answer.
Answer: D
A toolbox defines a curated set of tools once and exposes them through one MCP-compatible endpoint that agents on any MCP-compatible runtime, including LangGraph and Agent Framework, can consume. It centralizes authentication and guardrails, and tools can be added, removed, or reconfigured without changing agent code or redeploying. Per-agent tool configuration duplicates credentials, and a knowledge base provides grounding content, not tools.
Deep dive
How it works
A toolbox in Microsoft Foundry is a managed set of tools that you define once and expose through a single MCP-compatible endpoint. Agents consume that endpoint instead of carrying their own tool definitions and secrets.
Why it fits all twelve agents:
- Any runtime. Toolboxes are created in Foundry but aren't limited to Foundry agents. Any MCP-compatible runtime or client can use one, including custom agents built with Microsoft Agent Framework, LangGraph, or your own code.
- Centralized authentication. The toolbox handles credential injection, token refresh and policy enforcement at runtime, so consuming agents don't manage per-tool credentials.
- Governance in one place. You apply guardrails to tool inputs and outputs at the toolbox level.
- Change without redeploying. Because a toolbox is a managed resource, you can add, remove or update tools without changing agent code. With versioning, you promote a new default version and every agent that points to the toolbox picks it up.
MCP servers, OpenAPI tools and Azure AI Search are all on the toolbox's supported-tools list.
The other options, and when you'd use them
- Function calling in each agent, with copied code: Function calling is client-side execution and isn't a toolbox tool. Copying code into every app is the duplication a toolbox exists to remove.
- A Foundry IQ knowledge base with the tool definitions: A knowledge base is made of knowledge sources, connections to data stores, for grounding answers. It doesn't host or govern tools.
- Each tool directly on every agent with its own API keys: This is the problem the toolbox overview describes: credentials are duplicated and each agent manages its own secrets and token refresh.
Exam tip
"Same tools for many agents, any framework, central credentials and guardrails, change without redeploying" = Foundry toolbox behind one MCP endpoint.
References
Microsoft Learn references
Last reviewed: