Toronto Data Science & Big Data Meetup

Topic: Build Tool-Enabled AI Agents with MCP

Description: Hands-on introduction to Agentic AI using MCP and simple evaluation techniques.

Agentic AI is the next step beyond Chatbots: instead of only generating text, an agent can decide what to do, call tools, and verify outputs. In this 1-hour beginner-friendly hands-on session, we’ll build a simple “Study Assistant Agent” that uses the Model Context Protocol (MCP) to connect to tools in a clean, standardized way.

What you’ll build (live):
A small AI agent that:

• Routes a question to the right tool
• Uses MCP to call Notes Search and Calculator tools
• Adds a simple self-check (evaluation) step before finalizing the answer

What you’ll learn:

• What “Agentic AI” really means (decision → tool → check → answer)
• What MCP is and why it matters for real tool-enabled AI systems
• How to design tool interfaces and agent routing logic
• A simple, practical approach to evaluations (groundedness and correctness checks)

Prerequisites:

• Laptop recommended
• Basic Python helps (starter code provided)
• Optional: API key for an LLM (or follow along with provided mock setup)