Skip to main content
This workshop combines two cutting-edge technologies to transform how you build AI applications:
  • Model Context Protocol (MCP) — an open standard for seamless AI-tool integration
  • AI Toolkit for Visual Studio Code (AITK) — Microsoft’s powerful AI development extension
By the end of the workshop, you’ll have built production-ready MCP servers and integrated them with AI agents — all inside VS Code.

Technology stack

Model Context Protocol

MCP is the “USB-C for AI” — a universal standard connecting AI models to external tools and data sources. One protocol, infinite possibilities.
  • Standardized integration via stdio and SSE transport
  • Flexible local and remote server architectures
  • Built-in security and reliability

AI Toolkit for VS Code

Microsoft’s flagship AI development extension transforms VS Code into an AI development environment.
  • Model Catalog with 100+ models from Azure AI, GitHub, Hugging Face, and Ollama
  • Local inference with ONNX-optimized CPU/GPU/NPU execution
  • Visual Agent Builder with MCP integration
  • Real-time testing playground

Prerequisites

ComponentRequirementNotes
Operating SystemWindows 10+, macOS 10.15+, LinuxAny modern OS
Visual Studio CodeLatest stable versionRequired for AITK
Node.jsv18.0+ and npmFor MCP server development
Python3.10+Optional for Python MCP servers
Memory8 GB RAM minimum16 GB recommended for local models
Recommended VS Code extensions:
  • AI Toolkit (ms-windows-ai-studio.windows-ai-studio)
  • Python (ms-python.python)
  • Python Debugger (ms-python.debugpy)
  • GitHub Copilot (GitHub.copilot) — optional but helpful
Optional tools:
  • uv — modern Python package manager
  • MCP Inspector — visual debugging tool for MCP servers
  • Playwright — for web automation examples in Lab 2

Workshop labs

1

Lab 1: AI Toolkit Fundamentals (15 min)

Install and configure AI Toolkit for VS Code, then build your first AI agent.What you’ll do:
  • Install and configure AI Toolkit
  • Explore the Model Catalog (100+ models from GitHub, ONNX, OpenAI, Anthropic, Google)
  • Master the Interactive Playground for real-time model testing
  • Build your first AI agent with Agent Builder
  • Evaluate model performance with built-in metrics (F1, relevance, similarity, coherence)
  • Learn batch processing and multi-modal support capabilities
Learning outcome: A functional AI agent with a comprehensive understanding of AITK capabilities.Start Lab 1
2

Lab 2: MCP with AI Toolkit Fundamentals (20 min)

Connect MCP servers to AI Toolkit and build a browser automation agent.What you’ll do:
  • Master MCP architecture and concepts
  • Explore Microsoft’s MCP server ecosystem
  • Build a browser automation agent using the Playwright MCP server
  • Integrate MCP servers with AI Toolkit Agent Builder
  • Configure and test MCP tools within your agents
  • Export and deploy MCP-powered agents for production use
Learning outcome: A deployed AI agent supercharged with external tools through MCP.Start Lab 2
3

Lab 3: Advanced MCP Development (20 min)

Create custom MCP servers using AI Toolkit and the latest Python SDK.What you’ll do:
  • Create custom MCP servers using AI Toolkit
  • Configure and use the MCP Python SDK (v1.9.3)
  • Set up and use MCP Inspector for debugging
  • Build a Weather MCP Server with professional debugging workflows
  • Debug MCP servers in both Agent Builder and Inspector environments
Learning outcome: Custom MCP servers you can develop and debug with modern tooling.Start Lab 3
4

Lab 4: Practical MCP Development — GitHub Clone Server (30 min)

Build a real-world GitHub Clone MCP server for development workflows.What you’ll do:
  • Build a production-ready GitHub Clone MCP server
  • Implement smart repository cloning with validation and error handling
  • Create intelligent directory management with VS Code integration
  • Use GitHub Copilot Agent Mode with custom MCP tools
  • Apply production-ready reliability and cross-platform compatibility
Learning outcome: A production-ready MCP server that streamlines real development workflows.Start Lab 4

Real-world applications

The patterns you practice in these labs apply directly to enterprise use cases:
  • Smart repository management — AI-driven code review and merge decisions
  • Intelligent CI/CD — automated pipeline optimization based on code changes
  • Issue triage — automatic bug classification and assignment
  • Intelligent test generation — create comprehensive test suites automatically
  • Visual regression testing — AI-powered UI change detection
  • Performance monitoring — proactive issue identification and resolution
  • Adaptive ETL processes — self-optimizing data transformations
  • Anomaly detection — real-time data quality monitoring
  • Intelligent routing — smart data flow management
  • Context-aware support — AI agents with access to full customer history
  • Proactive issue resolution — predictive customer service
  • Multi-channel integration — unified AI experience across platforms

Skill mastery checklist

After completing all four labs, verify your skills: Core competencies:
  • MCP protocol architecture and implementation patterns
  • AI Toolkit proficiency for rapid development
  • Custom MCP server development, deployment, and maintenance
  • Tool integration — connecting AI with existing development workflows
Technical skills:
  • Set up and configure AI Toolkit in VS Code
  • Design and implement custom MCP servers
  • Integrate GitHub Models with MCP architecture
  • Build automated testing workflows with Playwright
  • Deploy AI agents for production use
  • Debug and optimize MCP server performance
Advanced capabilities:
  • Architect enterprise-scale AI integrations
  • Implement security best practices for AI applications
  • Design scalable MCP server architectures
  • Create custom tool chains for specific domains

Additional resources

MCP Specification

Full protocol specification (2025-11-25)

AI Toolkit GitHub Repository

Source code, issues, and samples for AITK

Sample MCP Servers

Community-maintained MCP server collection

OWASP MCP Top 10

Security best practices for MCP deployments

Next: Hands-On Labs

Continue to the 13-lab PostgreSQL integration capstone

Back: Case Studies

Review real-world MCP implementations

Build docs developers (and LLMs) love