Make your AI agent actually do things — tools, MCP, and what breaks

An AI agent that can only talk is a demo. One that can read files, call APIs and change things is a tool. We're working on the gap. Grand Rapids, all levels.

Second session, new venue: City Built Brewing downtown. Bring a laptop if you want to build along, or just come and watch.

WHAT WE'RE DIGGING INTO
Giving an agent real tools — file access, shell, web, APIs. How MCP (Model Context Protocol) fits in and why it caught on. Where tool-calling reliably falls apart, and the guardrails that help. I'll show what's working and what's still defeating me in Digi-Bot.

WHO THIS IS FOR
Developers using Claude Code, Cursor, Copilot or the APIs directly. Anyone who's wired up an agent and watched it confidently do the wrong thing. Beginners welcome — questions are what make this useful.

HOW THE EVENING GOES
6:30 — grab a beer, find the table, introductions
6:50 — Digi-Bot's tool layer: what works, what doesn't
7:20 — open floor: bring a problem or something to show
8:30 — whoever's still talking keeps talking

No cover. Food and drinks are on you. City Built has real tables and it's quiet enough to actually hear each other.