# interview-coach-agent-framework

**Repository Path**: binghe021/interview-coach-agent-framework

## Basic Information

- **Project Name**: interview-coach-agent-framework
- **Description**: No description available
- **Primary Language**: C#
- **License**: MIT
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2026-08-20
- **Last Updated**: 2026-08-20

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# Interview Coach with Microsoft Agent Framework

An AI-powered interview coach that shows how to wire up [Microsoft Agent Framework](https://aka.ms/agent-framework), [Model Context Protocol (MCP)](https://modelcontextprotocol.io), and [Aspire](https://aspire.dev) 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, Azure OpenAI, GitHub Copilot)
- Deploying to Azure with `azd up`

See [learning objectives](docs/LEARNING-OBJECTIVES.md) for the full breakdown.

## Architecture

![Overall architecture](./assets/architecture.png)

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 Foundry or Azure OpenAI

See [architecture overview](docs/ARCHITECTURE.md) for how the pieces fit together.

## Prerequisites

- [.NET 10 SDK](https://dotnet.microsoft.com/download/dotnet/10.0) or later
- [Visual Studio 2026](https://visualstudio.microsoft.com/downloads/) or [VS Code](https://code.visualstudio.com/download) + [C# Dev Kit](https://marketplace.visualstudio.com/items?itemName=ms-dotnettools.csdevkit)
- [Azure Subscription](https://azure.microsoft.com/free)
- [Microsoft Foundry](https://ai.azure.com)

See [LLM provider options](docs/providers/README.md) for alternatives.

## Getting Started

### 1. Clone Repository

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

### 2. Configure Microsoft Foundry

1. Create a new Microsoft Foundry project on Foundry Portal or command line.

   See [Foundry setup guide](docs/providers/MICROSOFT-FOUNDRY.md) for details.

### 3. Store Credentials

Use .NET user secrets to keep credentials secure:

```bash
dotnet user-secrets --file ./apphost.cs set MicrosoftFoundry:Project:Endpoint "{{MICROSOFT_FOUNDRY_PROJECT_ENDPOINT}}"
dotnet user-secrets --file ./apphost.cs set MicrosoftFoundry:Project:ApiKey "{{MICROSOFT_FOUNDRY_API_KEY}}"
```

### 4. Run the Application

Start all services with .NET Aspire:

```bash
aspire run --file ./apphost.cs
```

**What happens next:**

1. Open Aspire Dashboard (URL shown in terminal output).
1. All services start (Agent, WebUI, MCP servers, Cosmos DB).
1. Look for ✅ "Running" status on all resources.
1. Click the **webui** endpoint to open the interview coach.

### 5. Deploy to Azure

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

```bash
# Login to Azure
azd auth login

# Provision resources and deploy
azd up
```

### 6. Clean Up Resources

When finished, remove all Azure resources:

```bash
azd down --force --purge
```

## Next Steps

### Learn

- [Learning objectives](docs/LEARNING-OBJECTIVES.md)
- [Architecture overview](docs/ARCHITECTURE.md)
- [Tutorials](docs/TUTORIALS.md)
- [FAQ](docs/FAQ.md)

### Alternative LLM providers

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

- [Azure OpenAI](docs/providers/AZURE-OPENAI.md) — direct AOAI integration
<!-- - [GitHub Copilot](docs/providers/GITHUB-COPILOT.md) — local dev with Copilot SDK -->

### Alternative agent mode

The default is `LlmHandOff`, but you can also use:

- [`Single`](docs/MULTI-AGENT.md#mode-1-single-agent) - single-agent mode
<!-- - [`CopilotHandOff`](docs/MULTI-AGENT.md#mode-3-multi-agent-handoff-gitHub-copilot) - multi-agent mode with GitHub Copilot -->

## Additional Resources

### Microsoft Foundry

- [What is Microsoft Foundry?](https://learn.microsoft.com/azure/ai-foundry/what-is-foundry?view=foundry)
- [Foundry Agent Service](https://learn.microsoft.com/azure/ai-foundry/agents/overview?view=foundry)

### Microsoft Agent Framework

- [Framework Documentation](https://aka.ms/agent-framework)
- [Multi-Agent Orchestration](https://learn.microsoft.com/agent-framework/user-guide/workflows/orchestrations/overview)
- [AG-UI Protocol](https://docs.ag-ui.com/introduction)

### Model Context Protocol

- [MarkItDown MCP Server](https://github.com/microsoft/markitdown/tree/main/packages/markitdown-mcp)
- [MCP Specification](https://modelcontextprotocol.io)
- [MCP Server Registry](https://github.com/modelcontextprotocol/servers)

### Aspire

- [Aspire Documentation](https://aspire.dev)
- [Integrations](https://aspire.dev/integrations/overview/)
- [Deployment](https://aspire.dev/deployment/overview/)

## Contributing

We welcome contributions! Please see [CONTRIBUTING.md](docs/CONTRIBUTING.md) for guidelines.

## License

This project is licensed under the MIT License - see [LICENSE.md](LICENSE.md) for details.

---

Built by the CoreAI DevRel team | Questions? Check the [FAQ](docs/FAQ.md) or open an [issue](../../issue).