About this project

The repository is a structured learning resource rather than a single executable product. It introduces MCP as a standardized way for AI applications and clients to call tools and services exposed by MCP servers, using beginner-friendly explanations before progressing to implementation. The examples support .NET/C#, Java, TypeScript, JavaScript, Rust, and Python, beginning with calculator-style servers and clients and moving to LLM-connected applications. The curriculum is organized into modules. Foundation modules cover MCP introduction, core concepts, and security, including threat guidance and authorization patterns. The getting-started section walks through environment setup, first server and client, LLM integration, VS Code usage, stdio and HTTP streaming, Microsoft Foundry Toolkit, testing, deployment, RBAC and simple authentication, MCP hosts, Inspector, sampling, and MCP Apps. Later modules address practical SDK use, pagination, debugging, reusable prompt templates, multimodal scenarios, OAuth2, routing, scaling, advanced security, web search, real-time streaming and search, Microsoft Entra ID, Foundry integration, context engineering, custom transports, protocol features, and multi-agent reasoning examples. A dedicated 13-lab section guides learners through a PostgreSQL-integrated MCP server, covering architecture, multi-tenancy and row-level security, Docker and Azure setup, schema design, FastMCP implementation, database query tools, semantic search with Azure OpenAI and pgvector, testing, VS Code integration, Azure Container Apps deployment, monitoring with Application Insights, and production hardening. Additional material covers community contribution, adoption lessons, best practices, case studies, workshops, and Copilot App tooling. The README links official MCP documentation, the specification, SDK repositories, and community discussions. It identifies the taught protocol revision as 2026-07-28, including stateless requests and the Extensions framework, while noting that some examples remain versioned to 2025-11-25 while SDK support catches up. The repository also includes automated translations into more than 50 languages and documents sparse-checkout commands for cloning the core course without translation assets.