designing-real-world-ai-agents-workshop

Why we included this project

This workshop, presented at AI Engineering Conference Europe, walks you through building a multi-agent system from start to finish rather than stopping at the concept. The finished code includes two MCP servers: one for grounded deep research that pulls from search and YouTube transcripts, and one that drafts LinkedIn posts through a generate, review, edit loop. Both expose their tools and prompts over the Model Context Protocol, so you can plug them into a harness like Claude Code or Cursor. You have three ways in: watch the two-hour recording with slides, run the code in about half an hour to see a research brief and a scored post come out the other end, or rebuild the skeleton from pre-groomed tickets with agentic coding. Tool-use agents, structured LLM output, and MCP-server design all show up in a system that actually runs and produces artifacts, which is what sticks.