BEKADANAI & .NET Engineering
Join the waitlist

VIDEO COURSE

Build MCP Servers in C#: Connect AI to Your .NET Apps

Take the .NET app you already have and let AI use it safely: tools on your real services, every action audited, scoped by role, tested, traced and published.

  • ~10 h of video
  • 12 sections
  • 47 videos
  • .NET 10
  • MCP 2026-07-28
  • runs on your machine, no paid API keys

By the end, you can put an MCP server in front of any ASP.NET Core app, connect it to VS Code, Claude Desktop or your own C# client, and say exactly what the AI is allowed to do, as whom.

See the curriculum ↓
Build MCP Servers in C#: Connect AI to Your .NET Apps: cover

The problem this course solves

A hello-world MCP server takes ten minutes. Then you point it at the app your company actually runs. Does the tool skip your business rules because it talks to the DbContext? Why does the model guess a ticket number instead of searching? Who did the AI act as when it closed 20 tickets? Can a customer's ticket text tell the assistant what to do? What breaks when a second instance starts?

This course answers those questions on a real .NET app, with Inspector, tests and traces instead of hope.

What makes this course different

  1. 01

    You connect an app that already works.

    SupportDesk has services, business rules, 500 seeded tickets and a web UI. The MCP layer only calls those services. It never touches the DbContext.

  2. 02

    Every line explained.

    Each code video says what a line does and why it comes before or after the others, and says so when the order does not matter.

  3. 03

    You watch the AI change real data.

    The Activity page shows every write the AI makes, live, with the user and the client that made it.

  4. 04

    Built for the current spec.

    MCP 2026-07-28 removed sessions and the initialize handshake. Most tutorials still teach session code that no longer applies. This course uses SDK 2.x and stateless HTTP from the start.

What you will learn

  • Write tool names, descriptions and parameters the model calls correctly, and return errors it can recover from
  • Put MCP on top of your existing services, with search, paging, validation and a dry run before every bulk action
  • Expose resources and prompts users pick in the host: an SLA policy, ticket documents, a triage workflow
  • Move from stdio to stateless Streamable HTTP, Docker and two instances behind nginx
  • Protect the server with JWT bearer tokens and per-tool scopes, and limit what prompt injection in customer data can do
  • Build your own C# client on a local Ollama model with tool approval, then test, trace and publish the server to NuGet and the MCP Registry

Curriculum

  • Waitlist open
  • 12 sections
  • 47 videos + 1 text lesson
  • ~10 h of video

Lessons 1.1 and 1.3 and all of Section 2 are free previews (~29 min).

  1. SECTION 01Free preview

    Welcome & Setup

    Get SupportDesk running and see where you are going.

    3 videos + 1 text lesson · ~12 min

    After this section you can

    • Run SupportDesk with Docker and .NET 10
    • Reset to the same data as the videos at any time
    • See what the finished server does in VS Code

    Lessons

    1. 1.1What You'll BuildFree preview~4 min
    2. 1.2Tools You Need~3 min
    3. 1.3Running the SupportDesk StarterFree preview~5 min
    4. 1.4TextSetup Troubleshootingtext

  2. SECTION 02Free preview

    MCP Fundamentals

    What MCP is, who talks to whom, and what changed this year.

    5 videos · ~20 min

  3. SECTION 03

    Your First MCP Server

    From an empty project to a tool three hosts can call.

    5 videos · ~1 h 8 min

  4. SECTION 04

    Tools in Depth

    The model only sees your names and descriptions. Make them count.

    6 videos · ~1 h 27 min

  5. SECTION 05

    Connecting Your Existing App

    Tools are thin. Services do the work.

    6 videos · ~1 h 29 min

  6. SECTION 06

    Resources & Prompts

    Not everything should be a tool.

    4 videos · ~1 h 2 min

  7. SECTION 07

    HTTP Transport & Deployment

    One server for everyone. Any instance answers any request.

    5 videos · ~1 h 1 min

  8. SECTION 08

    Security

    Prompts ask. Servers decide.

    4 videos · ~55 min

  9. SECTION 09

    Building an MCP Client

    Your own host, on a local model.

    3 videos · ~42 min

  10. SECTION 10

    Production Readiness

    Tested, traced and installable with one command.

    3 videos · ~57 min

  11. SECTION 11

    Beyond the Basics

    Work that takes minutes, and answers that need a screen.

    2 videos · ~45 min

  12. SECTION 12

    Wrap-Up

    What you built, and how to do the same in your own app.

    1 video · ~5 min

The project: SupportDesk

SupportDesk is the help desk of Brightline, a made-up SaaS company: tickets, customers, products, agents, SLAs and an activity log. It is built with .NET 10, ASP.NET Core, EF Core 10 and PostgreSQL 17, and starts with 500 seeded tickets, the same data you see in the videos. You get it running in the first section. The course adds a separate SupportDesk.Mcp project that calls the same application services as the web app, so the AI follows the same rules people do.

  • Tool · readget_ticketOne ticket with its comments; also shown as an interactive card in hosts that support MCP Apps
  • Tool · readget_customerA customer by email or id, with recent tickets
  • Tool · readsearch_ticketsFilters by keyword, product, agent, status, priority and SLA, paged
  • Tool · readget_ticket_statsTicket counts for the last N days, grouped by product, priority or status
  • Tool · writeadd_commentAdds a public reply or an internal note
  • Tool · writeupdate_ticket_statusChanges status, following the same rules as the web app
  • Tool · writeassign_ticketAssigns an agent by name or email, or unassigns
  • Tool · writecreate_ticketOpens a ticket; SLA and number are set by the server
  • Tool · bulkclose_stale_ticketsPreviews first, closes only a confirmed count, admins only
  • Tool · long-runningexport_ticket_reportBuilds a CSV report as a Task with progress, admins only
  • Resourcesupportdesk://kb/sla-policy, supportdesk://kb/escalation-guideDocuments the user attaches as context
  • Resource templatesupportdesk://tickets/{ticketNumber}, supportdesk://customers/{customerId}Tickets and customers as documents
  • Prompttriage_ticket, draft_customer_reply, weekly_summaryWorkflows the user starts from the host

You connect it to VS Code with GitHub Copilot, Claude Desktop, MCP Inspector and SupportDesk Chat, a C# client you build yourself on a local model.

What you get

  • 47 videos (~10 h) in plain English with English subtitles, plus a setup troubleshooting guide
  • SupportDesk's full source code, with a git tag at the end of every section and a commit link under every video
  • The finished MCP server: 10 tools, 4 resources, 3 prompts and an MCP App, with Docker, nginx, tests and OpenTelemetry
  • SupportDesk Chat, your own C# MCP client with tool approval on a local model
  • README guides for connecting VS Code, Claude Desktop and Inspector, running over HTTP and Docker, and packaging and publishing

Requirements

  • C#, ASP.NET Core and EF Core experience
  • .NET 10 SDK, Docker Desktop, VS Code with C# Dev Kit, and Node.js 22.19+ for MCP Inspector
  • Ollama for Section 9, with a model that supports tool calling such as llama3.1:8b. A GPU is optional; it makes answers faster.
  • No paid services or API keys. GitHub Copilot Free and the free Claude Desktop are enough.

Who this course is for

FOR YOU IF YOU ARE

  • .NET developers who want AI assistants to use their app without bypassing its business rules
  • Team leads who must answer "what can the AI do in our system, as whom, and can we see it?"
  • Developers who built a demo MCP server and now need one that holds up with real users

NOT FOR

  • Developers new to C#, ASP.NET Core or EF Core
  • Anyone looking for RAG, agent orchestration or evals (those are in AI Engineering with .NET)

ABOUT

Taught by Burak Emre Kadan

I've spent 12 years building .NET backends — most recently leading teams that design distributed systems on Azure. My day-to-day has been the unglamorous parts that keep production running: retries and circuit breakers, idempotent consumers, outbox patterns, tracing with OpenTelemetry, and decisions written down as ADRs.

That's the lens I bring to AI. An LLM call is just another unreliable dependency — it needs the same engineering discipline as everything else in your system. This course teaches exactly that, in C#.

MORE ABOUT ME
Experience
12 years in C# and .NET
Focus
Distributed systems, reliability, observability
Format
A self-paced video course with working source code

FAQ

Do I need a paid AI subscription?

No. VS Code with GitHub Copilot Free, the free Claude Desktop, MCP Inspector and a local Ollama model cover every demo.

Do I need a GPU?

No. Only Section 9 runs a local model, and it works on a CPU, just slower.

Do I need a cloud account?

No. Everything runs on your machine with Docker. Section 8 shows what changes for a real identity provider.

Is this up to date with the latest MCP spec?

Yes. It uses MCP 2026-07-28 and the official C# SDK 2.x. Section 2 shows which session-based code from older tutorials no longer applies.

Why aren't sampling, roots and elicitation covered?

In SDK 2.x they are obsolete or limited on stateless HTTP. The course teaches what works on the current spec.

Will my data match the videos?

Yes. The seed data is fixed. Run the app with --reset-db at the start of each section and you see the same tickets.

Can I use this with my own app?

Yes. The MCP layer only calls application services, and the last lesson walks through the steps to do the same in your app.

How is this different from AI Engineering with .NET?

This course is about letting AI use your app through MCP. RAG, agent orchestration and evals are covered in AI Engineering with .NET.

WAITLIST

Hear first when Build MCP Servers in C# opens

Join the waitlist and you will get an email when the course opens.