🧠 Build an AI Backend: Tools, Memory & Prompts for Real-World LLMs

Welcome to the workshop! In this hands-on session, you'll learn how to build and deploy an MCP (Model Context Protocol) server — the backend brain behind AI clients like assistants, copilots, and autonomous agents.

Using the open-source TypeScript MCP SDK and Claude/Cloudflare Playground, you’ll create a system that gives an LLM memory, tool access, and structured prompts — and deploy it live using Cloudflare.

No prior AI experience needed — just high-level JavaScript/TypeScript, and curiosity!

By the end of this workshop, you will:

• Understand what an MCP Server is and how it powers AI systems

• Use the MCP SDK to:

  • Register and expose tools (functions) to an AI model

  • Manage context and memory for stateful conversations

  • Orchestrate prompts and responses between clients and Claude

• Run and test your MCP server locally

• Deploy your MCP server to Cloudflare using Workers or Pages Functions so others can use it too

This workshop is designed for:

• Engineers curious about AI infrastructure

• Folks building AI-powered assistants, bots, or internal tools

• DevOps/Platform engineers interested in tool orchestration

• Anyone who wants to get hands-on with real LLM systems (beyond chatbots)

All tools are free, open source, or have free tiers. We'll support you with setup.

You’ll need:

• A personal GitHub account

• A working Node.js v18+ environment

• An account with Claude or mcp client of choice

• A Cloudflare account (free tier is fine)

👉🏾Full set-up instructions are here

• A fully working AI backend with:

  • Custom tools (functions) that your AI can call

  • Memory for storing state or chat history

  • Prompt logic that guides Claude’s behavior

• A live, working Cloudflare-hosted MCP server accessible from the web

• A GitHub repo of your custom, live-deployed MCP server

• Experience building modern AI infrastructure with open source tools

• A clearer understanding of AI backend architecture and edge deployments

• Confidence to keep building LLM-powered assistants, agents, or copilots

• Real experience with Cloudflare’s developer platform (Workers, Pages, etc.)