Building Next Gen Agents with MCP
summary
Builds a working Model Context Protocol tool chain: a Chess MCP server that accepts algebraic notation, uses chess.js for rule enforcement, and stores per-session state as JSON on disk so games persist across turns. Client side uses Nanobot as the MCP client and MCP‑UI to surface an ASCII view and a drag-and-drop chessboard iframe. Runs on Node.js 20+ with TypeScript and an OpenAI key, communicates over stdio or streamable HTTP, and demos full round trips where the AI legally moves and the board updates.
AI-written summary · about
chapters (10)
- 0:00Introduction to MCP
- 4:47Setting up the environment
- 9:12Building the MCP server
- 18:27Running the server with Nanobot
- 31:15Adding chess functionality
- 41:03Implementing game state persistence
- 51:40Enhancing UX with starter messages
- 57:39Introducing MCP UI
- 1:04:00Adding graphical chessboard
- 1:09:14Q&A and next steps
related talks
30mComposable AI: Building Next-Gen AI Agents with MCP
16mDriving agentic innovation w/ MCP as the backbone of tool-aware AI
45mAgents, Tools, and MCP, oh my! Next-level AI concepts for developers
33mThe Model Context Protocol: Connecting AI to Everything
32mFrom Monolith to AI Agent Modernizing Java Systems with MCP
22mNext-Gen AI Tooling with MCP Servers in Go
50mBeyond local tools: Dive into the Model Context Protocol (MCP)
12mMCP and the Dawn of Agentic Intelligence