Module 1 · The protocol and its mental model · Lesson 1

MCP and REST: two mental models

Why MCP exists, how its design differs from REST's constraints and where the 2026 revision moved towards them, and what wrapping a REST API one endpoint per tool costs an agent.

75 min 2 videos 3 exercises 8-question quiz Answers

Setup for the exercises:Install Claude Code Node.js, pnpm and TypeScript MCP servers in Claude Code The Larchmere Desk mock API

By the end of this lesson you can

Explain the integration problem MCP solves and the roles of host, client and server
Compare REST and MCP on who decides to call, granularity, discovery, state, direction, errors, caching, safety and authorization, separating sourced facts from inference
Describe which of REST's constraints the stateless 2026-07-28 revision adopted, and why MCP is still RPC
Choose between an MCP server, a REST API, a command-line tool and a skill for a given consumer
Measure an endpoint-shaped MCP server's definition size, calls and tokens on agent tasks, as a baseline to improve on

Larchmere's scripts, status page and integrations have used the Larchmere Desk REST API for years. Now three agents need Desk too: Claude Code for engineers, a triage agent, and Claude for the support team. The obvious move is to wrap each REST endpoint as an MCP tool, and most MCP servers are built exactly that way. Is that the right shape? This lesson compares the mental models behind REST and MCP, shows where the current MCP revision borrowed REST's constraints and where it differs on purpose, and ends with you building the obvious server and measuring it, so later lessons have a baseline to beat.

MCP and REST: two mental models

Why MCP exists, how its design differs from REST's constraints and where the 2026 revision moved towards them, and what wrapping a REST API one endpoint per tool costs an agent.

MCP and REST: two mental models

MCP and REST: two mental models

Sign in to keep reading.

Sign in to keep reading.