The Put That Sh*t On Agent: Building Multi-Agent AI Systems That Actually Talk to Each Other*
summary
A working multi‑agent wardrobe app connects a central orchestrator to specialist agents via the Agent2Agent (A2A) protocol, uses MCP to fetch live weather and calendar data, Conduit for real‑time pipelines, and runs on Cloud Run; it builds outfits, performs a virtual try‑on, and enforces an ontology rulebook that validates proposals and triggers a single‑step repair loop (swap shoes) before falling back to deterministic rules. Iteration takes months; system reduces bad suggestions without disabling creative outputs.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 1:57MCP and A2A protocols
- 3:21Outfit recommendation system
- 5:26AI wardrobe assistant demo
- 6:29Hallucinations as a feature
- 9:00Ontologies for data rules
- 12:55Facts from clothing data
- 14:20Validating outfit suggestions
- 16:30Final system demo
- 18:59Improvements and challenges
- 20:23When to use ontologies
- 22:29Real-world applications
- 22:50Models propose contracts
- 23:30Q&A
related talks
41mPatterns & Practices for building Multi-Agent Systems
45mAn Agent Walks Into a Bar... and Meets Another Agent: Multi-Agent Systems in Action
49mAI Agents: Collaboration in Real-Life Applications
32mBuilding Multi-Agent Systems: From Concepts to Collaboration
40mAI Agents in Practice
26mBuilding Agentic Workflows: Patterns for Orchestrating Intelligent Systems.
27mMulti-Agent AI: Transforming Business Operations and Value Creation
01h 04mAgentic AI: Build a Multi-Agent Application with CrewAI