Your first AI agent is only a few lines of Python away

Microsoft Developer demonstrates how to turn a deployed model into a reusable AI agent using the Microsoft Foundry SDK in Python, including adding instructions, running a multi-turn conversation, and wiring in MCP tools so the agent can take actions while maintaining context.

Overview

The video walks through creating a simple “cupcake-ordering” prompt agent with the Microsoft Foundry SDK.

Links

• Azure AI Foundry portal: https://aka.ms/foundry-portal
• Inside Microsoft Foundry playlist: https://aka.ms/InsideMicrosoftFoundryPlaylist

What the session covers

• Build a cupcake-ordering agent
  • Starts from a deployed model and wraps it as an agent.
• Set up Microsoft Foundry
  • Uses the Foundry environment and SDK to get ready to build.
• Create and test the agent
  • Runs an initial interaction to validate the agent behavior.
• Add instructions and MCP tools
  • Adds instruction text that defines how the agent should behave.
  • Introduces MCP tools so the agent can use tools during execution.
• Complete a cupcake order
  • Demonstrates a multi-turn conversation where the agent maintains context.
• Recap and key takeaways
  • Summarizes how a basic model deployment becomes an agent you can extend.

Timeline

• 0:00 — Build a Cupcake-Ordering Agent
• 0:36 — Set Up Microsoft Foundry
• 1:41 — Create and Test the Agent
• 2:32 — Add Instructions and MCP Tools
• 3:20 — Complete a Cupcake Order
• 4:27 — Recap and Key Takeaways