A collection of C# demos showcasing practical AI agent patterns, including group-chat orchestration, Model Context Protocol (MCP) integration, and Windows Event Log analytics.

• .NET 10 SDK (preview)
• Windows host for Windows Event Log–based samples
• Optional administrator rights for richer System log access in Event Log–based samples
• Azure OpenAI access (required for JokeAgentsDemo)
• Additional project-specific requirements are noted below and in each project’s README

• Open the solution: AIForCSharpDev.slnx
• Or run projects individually with dotnet run --project <ProjectName>

• Path: HelloAgent/
• Summary: Simple “hello world” agent sample to get started quickly.
• Setup/Run: See project directory for details.

• Path: JokeAgentsDemo/
• Summary:
  • Demonstrates a two-agent “creator and critic” workflow using the Microsoft Agent Framework (MAF)
  • Group chat orchestration with a custom quality gate that stops when the critic approves or rates highly
  • Browser UI with both streaming (SSE) and classic result views
  • REST and SSE endpoints for automation and integrations
• Prerequisites:
  • .NET 10 SDK (preview)
  • Azure OpenAI with a deployed chat-completion model
  • Local Azure authentication (az login)
• Configure:
  • Update Azure OpenAI settings in Program.cs:

    var endpoint = new Uri("https://<your-resource>.cognitiveservices.azure.com/");
    var credential = new DefaultAzureCredential();
    string deploymentName = "<your-deployment-name>";

• Run:

  dotnet run --project JokeAgentsDemo

  Open http://localhost:5000 and choose the streaming or classic experience.
• Endpoints:
  • GET / — Web UI
  • POST /api/jokes/create?topic=<topic> — One-shot workflow
  • GET /api/jokes/stream?topic=<topic> — Streaming via SSE
  • POST /agents/creator — Direct A2A to creator
  • POST /agents/critic — Direct A2A to critic
  • GET /agents/{agent}/.well-known/agent.json — Agent cards
  • GET /health — Health check
• More: See the project’s README at JokeAgentsDemo/README.md

• Path: EventLogMcp/
• Summary:
  • Native C# MCP server exposing Windows Event Log analytics as AI tools
  • HTTP MCP endpoint at http://localhost:5115/mcp
  • Tools:
    • get_startup_shutdown_events — Startup, shutdown, and unexpected shutdown events for a period
    • calculate_uptime — Aggregate and per-day uptime statistics
    • get_usage_summary — Compact, human-readable usage summary with stats
  • Automatic fallback from System log to Application log when admin rights are unavailable
  • JSON responses with camel-cased properties suitable for agents
• Requirements:
  • Windows host with access to the local Event Log
  • .NET 10 SDK (preview), targets net10.0-windows
  • Optional admin rights (for System log)
• Run:

  dotnet run --project EventLogMcp

  The server listens on http://localhost:5115/mcp.
• More: See EventLogMcp/README.md

• Path: ComputerUsageAgent/
• Summary:
  • Agent that connects to the MCP server from EventLogMcp to analyze Windows computer usage via Event Logs
• Run:
  1. Start the MCP server:

     cd EventLogMcp
     dotnet run

  2. Run the agent:

     cd ../ComputerUsageAgent
     dotnet run

• More: See ComputerUsageAgent/README.md

• Path: LocalOllamaAgent/
• Summary: Local model agent demo. See the project directory for setup and usage details.

AIForCSharpDev.slnx
ComputerUsageAgent/
EventLogMcp/
HelloAgent/
JokeAgentsDemo/
LocalOllamaAgent/
Setup/
LICENSE.txt
CHANGES.md
PROJECT-STRUCTURE.md
TEST_PLAN.md

MIT — see LICENSE.txt