mcp-for-beginners

Provides structured curriculum and working code examples for building MCP servers in six programming languages (Python, TypeScript, JavaScript, C#, Java, Rust) using language-specific SDKs (FastMCP for Python, native TypeScript/JavaScript, Spring AI for Java, etc.). Each language implementation follows the same protocol specification but leverages native idioms, async patterns, and ecosystem conventions, enabling developers to choose their preferred language while maintaining protocol compliance.

Teaches and demonstrates the complete lifecycle of MCP client-server communication: session initialization, capability negotiation, request routing, and graceful shutdown. Abstracts transport mechanisms (stdio, HTTP streaming, custom transports) behind a unified protocol layer, allowing clients to communicate with servers regardless of underlying transport. Includes patterns for connection pooling, error recovery, and message serialization/deserialization using JSON-RPC 2.0.

Teaches how to extend MCP servers to handle multimodal inputs (text, images, audio, video) and outputs, and how to engineer context for multimodal LLMs. Covers resource types for different media formats, streaming binary data over MCP, and optimization patterns for large media files (compression, chunking, lazy loading). Includes examples of image analysis tools, document OCR, and video processing integrated via MCP.

Demonstrates how to integrate web search capabilities and external data sources (APIs, databases, knowledge bases) into MCP servers, enabling LLMs to access real-time information and enterprise data. Covers patterns for wrapping REST APIs as MCP tools, implementing search result ranking and filtering, caching external data, and handling rate limits and authentication for external services.

Teaches how to integrate databases into MCP servers with row-level security (RLS), multi-tenancy support, and secure data access patterns. Covers SQL query building with parameterization to prevent injection, connection pooling, transaction management, and authorization checks at the row level. Includes examples of integrating relational databases (PostgreSQL, SQL Server) and NoSQL databases (MongoDB) with MCP, with explicit patterns for enforcing tenant isolation and user-based access control.

Provides a four-phase, 11-module curriculum structure (Foundation, Building, Growth, Mastery) with progressive complexity, hands-on labs, and real-world case studies. Each module includes README documentation, working code examples in six languages, and practical exercises. Foundation phase covers protocol basics and security; Building phase teaches implementation; Growth phase covers practical patterns; Mastery phase addresses advanced topics (cloud integration, scaling, multimodal support). Case studies include Microsoft Learn Documentation MCP Server, Azure AI Travel Agents, and GitHub MCP Registry integration.

Provides curriculum and patterns for defining MCP resources (URIs, MIME types, content) and tools (function signatures via JSON Schema) with built-in validation. Resources are declared with URI templates and content types; tools are defined as JSON Schema objects with input/output specifications. The curriculum demonstrates how to validate incoming requests against schemas, handle schema evolution, and expose schema metadata to clients for capability discovery and type safety.

Demonstrates how to integrate MCP servers with LLM clients (OpenAI, Anthropic, local models) by injecting MCP resources and tool definitions into the LLM's context window. Teaches context engineering patterns: resource prefetching, tool ranking by relevance, token budget management, and dynamic context selection based on user queries. Includes examples of connecting MCP servers to Claude, GPT-4, and open-source models via standard LLM APIs.

Comprehensive curriculum covering AI-specific security threats in MCP deployments: prompt injection via tool outputs, privilege escalation through resource access, supply chain attacks via untrusted servers, and data exfiltration. Teaches authentication patterns (API keys, OAuth 2.0, mTLS, managed identity), authorization models (role-based access control, row-level security), and secure coding practices. Includes Microsoft Security Ecosystem integration (Azure AD, Managed Identity) and threat-specific mitigations.

Teaches how to integrate MCP servers into VS Code as extension providers, enabling in-editor access to tools and resources. Demonstrates the VS Code Extension API integration points, message passing between extension and MCP server, UI patterns for displaying tool results, and debugging MCP servers from within VS Code. Includes case study of Microsoft Learn Documentation MCP Server integrated into VS Code for in-editor documentation lookup.

Provides testing patterns and debugging strategies for MCP servers across all six supported languages. Includes unit testing frameworks (pytest for Python, Jest for TypeScript/JavaScript, xUnit for C#, JUnit for Java), integration testing with mock MCP clients, and debugging techniques (logging, breakpoints, message inspection). Teaches how to test transport layers, validate schema compliance, and simulate error conditions (network failures, timeouts, malformed requests).

Teaches deployment strategies for MCP servers across different environments: local development (stdio), cloud platforms (Azure, AWS), containerized deployments (Docker, Kubernetes), and serverless functions. Covers load balancing, intelligent routing (routing requests to appropriate server instances based on capability), connection pooling, and horizontal scaling. Includes patterns for multi-instance deployments with shared state management and health checking.

Teaches how to implement real-time, bidirectional communication in MCP using streaming transports and notification mechanisms. Covers server-to-client notifications (resource updates, tool execution progress), client-initiated streaming (long-running operations), and event subscription patterns. Includes examples of WebSocket-based streaming, Server-Sent Events (SSE), and custom transport implementations for real-time use cases.

Demonstrates integration of MCP servers with Azure AI Foundry, including keyless authentication via Azure Managed Identity, integration with Azure OpenAI and other Azure AI services, and deployment patterns for Azure Container Instances, App Service, and Kubernetes. Teaches how to use Managed Identity for secure, credential-less authentication between MCP servers and Azure services, eliminating the need for API keys or connection strings.