Interview Coach with Microsoft Agent Framework

An AI-powered interview coach that shows how to wire up Microsoft Agent Framework, Model Context Protocol (MCP), and Aspire into a working application you can deploy.

What you'll learn

This sample covers the patterns you'd need for a real agent deployment:

• Building AI agents with Microsoft Agent Framework
• Multi-agent handoff orchestration — single agent vs. 5 specialized agents
• Model Context Protocol (MCP) for adding tools without touching agent code
• Running multiple services together with Aspire
• Keeping conversation state across sessions
• Swapping LLM providers (Microsoft Foundry and GitHub Copilot)
• Deploying to Azure with azd up

See learning objectives for the full breakdown.

Architecture

The app is split into a few services:

• Aspire orchestrates everything (service discovery, health checks, config)
• WebUI is a Blazor chat interface
• Agent runs the interview logic via Microsoft Agent Framework
• MCP Servers handle document parsing (MarkItDown) and session storage (InterviewData)
• LLM Provider talks to Microsoft Foundry or GitHub Copilot

See architecture overview for how the pieces fit together.

Prerequisites

• .NET 10 SDK or later
• Visual Studio 2026 or VS Code + C# Dev Kit
• Docker Desktop or equivalent container runtime

Microsoft Foundry also requires an Azure subscription and the Azure CLI. GitHub Copilot requires a Copilot-enabled account and GitHub authentication. See LLM provider options.

Getting Started

1. Clone Repository

git clone https://github.com/Azure-Samples/interview-coach-agent-framework.git
cd interview-coach-agent-framework

2. Choose an LLM provider

Microsoft Foundry is the default. Aspire provisions its resource and gpt-5-mini deployment when the app starts. See the Foundry setup guide, or use GitHub Copilot without provisioning an Azure model.

3. Configure authentication

For Microsoft Foundry, sign in with the Azure CLI:

az login

The agent uses DefaultAzureCredential: Azure CLI credentials locally and managed identity when deployed. For GitHub Copilot, follow the GitHub authentication setup.

4. Run the Application

Start all services with .NET Aspire:

aspire start --apphost ./apphost.cs

What happens next:

1. Open the Aspire Dashboard from the URL printed in the terminal.
2. Wait for the services to report Running.
3. Open the webui endpoint.

5. Deploy to Azure

Deploy the entire application to Azure Container Apps with one command:

# Login to Azure
azd auth login

# Provision resources and deploy
azd up

6. Clean Up Resources

When finished, remove all Azure resources:

azd down --force --purge

Next Steps

Learn

• Learning objectives
• Architecture overview
• Tutorials
• FAQ

Alternative LLM providers

The default is Microsoft Foundry, but you can also use:

• GitHub Copilot — GitHub Copilot SDK integration

Alternative agent mode

The default is HandOff. You can switch to:

• Single - single-agent mode

Additional Resources

Microsoft Foundry

• What is Microsoft Foundry?
• Foundry models

Microsoft Agent Framework

• Framework Documentation
• Multi-agent orchestration
• AG-UI Protocol

Model Context Protocol

• MarkItDown MCP Server
• MCP Specification
• MCP Server Registry

Aspire

• Aspire Documentation
• Integrations
• Deployment

Contributing

We welcome contributions! Please see CONTRIBUTING.md for guidelines.

License

This project is licensed under the MIT License - see LICENSE.md for details.

---

Built by the CoreAI DevRel team | Questions? Check the FAQ or open an issue.