Build Hackathon: Ship an AI Agent in 4 Hours
Half-day hackathon where every team leaves with a working AI agent shipped to the web
A half-day hackathon format. Teams of 3-5 take one real problem from their work, decompose it into an agentic workflow, build with Claude Code (or a comparable agent CLI), and ship a working prototype to a live URL. The facilitator role shifts from teacher to co-pilot — Frank pairs with each team during the build, unblocks blockers, and pushes for shipped-over-polished. By the close, every team has a live link they can share.
Learning Objectives
Pick one real problem from the team's work that AI can materially help with
Decompose it into agent tools + memory + reasoning in a shared whiteboard
Build a functional prototype with Claude Code or equivalent
Deploy to a live URL (Vercel preview, Cloudflare Pages, or Railway)
Demo in a 4-minute slot to the room with a working link
Prerequisites
- Each team has one real problem to attack — bring a real use case, not a toy
- Laptop with Node.js 18+ and a git account
- API keys or cloud credits for at least one LLM (Claude, OpenAI, or Gemini)
- Comfort with terminal OR willingness to pair with a team member who is
Workshop Agenda
Teams map their problem into the agent primitives: what tools does the agent need, what memory does it keep, what reasoning loop does it run. Facilitator circulates with the 6-pillar CoE framework as a quick sanity check on governance and data.
Key Resources
Share this workshop
Get the Resource Pack
Receive the complete slide deck, handouts, and facilitator guide for this workshop.