Module 3 · Project 2: Company Tools as an MCP Server (Model Context Protocol)

OBJECTIVE: turn the shop's data and actions from Project 1 into an MCP server, so any MCP-compatible AI app (Claude Desktop, Claude Code, Cursor, VS Code, your own LangGraph agent) can use them safely — and learn MCP from basics to a remote server. WHAT WE COVER (in project order): • Why MCP: one standard way to connect AI apps to tools and data • MCP architecture: host, client, server; tools, resources and prompts • Building an MCP server in Python • Testing with the MCP Inspector and using it from a desktop AI app • Connecting MCP tools to your LangGraph agent • Local (stdio) vs remote (HTTP) servers and authentication basics • MCP security: tool poisoning, least privilege, approvals AFTER THIS MODULE YOU CAN: • Build, test and connect an MCP server • Reuse one set of tools across many AI apps and agents • Spot and prevent the main MCP security risks DATA: the shop database from Project 1. TOOLS: MCP Python SDK (FastMCP), MCP Inspector, langchain-mcp-adapters, an MCP host app.