Microsoft Developer leads a practical session on building an intelligent pizza ordering agent using Microsoft Foundry Agent Service and MCP integration, featuring custom tools and external data.

Build a Pizza Ordering Agent with Microsoft Foundry and MCP

Microsoft Developer presents a hands-on workshop for constructing domain-specific AI agents, illustrated with the Contoso PizzaBot example. The session progresses through the core steps of agent creation, utilizing Microsoft Foundry Agent Service and integrating with the MCP server for real-time menu and order management.

Key Topics Covered

• Agent Creation:

  • Step-by-step guide to building a basic AI agent.
  • Designing tailored system prompts for customized interaction.

• Domain-Specific Intelligence:

  • Leveraging retrieval-augmented generation (RAG) to answer questions from uploaded documents.
  • Ensuring agents access and process domain-specific knowledge efficiently.

• Custom Tool Integration:

  • Adding external functionalities such as a pizza calculator.
  • Demonstrating how agents can call application logic based on user requests.

• Live Integrations:

  • Connecting to an MCP (Microsoft Cloud Partner) server.
  • Managing dynamic menus and live order processing within the AI framework.

Practical Resources

• Workshop Lab Repo
• Join the Discussion on Discord
• Sign Up for Azure Free Trial
• On-Demand Session Recordings

What You’ll Learn

• Foundations of building AI agents with Microsoft Foundry
• Applying RAG for situational question answering
• Integrating custom tools and handling external data sources
• Achieving live updates and transaction flows with MCP APIs

Example: Contoso PizzaBot

The session demonstrates Contoso PizzaBot—a pizza ordering assistant capable of responding to user queries, performing order calculations, and fetching live menu data. The agent reacts to tailored prompts and employs RAG to match user requests with custom documentation.

The integration with an MCP server showcases real-time business logic and menu management, illustrating how enterprise applications can benefit from these AI-driven solutions.

Next Steps

• Follow the workshop to build your own domain-specific AI agent.
• Experiment with custom tool and data source integration.
• Continue learning via provided labs and community resources.

For continuing discussion and support, visit the Discord channel or review on-demand materials.