AI Technology Workshop Series: AI Agents with MCP

Course Description

Welcome to our immersive AI technology workshop series. During these sessions you will be introduced to new and established AI tools that will help you create and manipulate content in new and powerful ways. Each session is led by an industry expert who will guide you through the material and share its real-world implications.

Learning Outcomes  
By the end of the workshop, participants will be able to:

• Explain and build a working Model Context Protocol (MCP) server that exposes tools, data, and reusable prompts to any AI app.
• Create an AI agent that discovers and uses tools from multiple MCP servers to complete multi-step tasks.
• Test and debug servers and agents with hands-on tracing and a small evaluation set.
• Design tools that AI can use reliably, with clear descriptions, sensible scope, and helpful error handling.
• Defend agents against prompt injection, and add human approval for risky actions.

Topics Include

• Agents and MCP fundamentals: how the agent loop works, and why MCP is becoming the standard way to connect AI to tools and data
• Hands-on MCP server development: tools, resources, and prompts, built in Python from scratch
• Building agents with MCP: tool-use loops, multi-server orchestration, and state management
• Tool design best practices: writing agent-friendly tool names, descriptions, and error messages
• Agent safety and evaluation: prompt injection defenses, human-in-the-loop approvals, and testing agent behavior

Prerequisites: Knowledge of LLM and AI Agents.

Note: Students will need to have Google Cloud, GitHub, and Hugging Face accounts to participate in hands-on exercises. Students are expected to finish the assignment to be graded.