AI models are sycophants. Ask one to manage your infrastructure and it will confidently invent API responses, fabricate deployment results, and return whatever answer it thinks you want to hear. It is not lying, it is optimized to please, and it is very good at sounding right even when it is shooting darts in the dark.
This workshop introduces attendees to the Model Context Protocol (MCP) and demystifies why tool calling changes the equation. Participants will deploy MCP servers hands-on, then experience the before and after firsthand: the same model, the same prompt, completely different behavior once it has real tools to call instead of blank space to fill.
All lab elements run locally. Attendees can use an optional deployment of Ollama for a fully offline experience or connect API keys to the LLM of their choice, Anthropic, OpenAI, or Google Gemini. The github project will be available as a take home to continue learning after the conference.