MicrosoftMicrosoft Certified: Multi-Agent AI Solutions Expert2026-101.2 Specify technology components to include in a multi-agent solution
Build AI agents with portal and VS Code
Completed by opening the official lab
Why it matters
The same Foundry prompt agent is reached from the portal, the editor and code, and knowing which surface owns which step is part of choosing a development and hosting approach.
What to practise
Create an agent in the Microsoft Foundry portal with instructions, file search over an uploaded policy document and code interpreter over a CSV file. Test it under Prompt Agents in the Foundry Toolkit for VS Code, which some labels show by its earlier name, AI Toolkit. Then run a Python client that uses AIProjectClient, a conversation and the Responses API, and saves the charts the agent generates. The client names the agent through extra_body.agent_reference on the shared project endpoint, the Agents v1-era pattern that an agent-bound client replaces. The exercise notes that some of its features are in preview.
How it fits
This is the prompt-agent option from 1.2.1: Foundry hosts the agent and its tools, while your code only calls it. The editor tooling it introduces is the extension side of 1.2.9’s development environment.