hub

Model Context Protocol — Deep Dive

From protocol basics to production MCP servers. Each chapter: high-level overview + under-the-hood deep dive.
Co-Created by Kiran Shirol and Claude
Core Stack MCP JSON-RPC 2.0 Python SDK TypeScript SDK
home Learning Portal play_arrow Start Learning summarize Key Insights dictionary Glossary 10 chapters · Each with High Level + Under the Hood
Protocol

Architecture & Transports

What MCP is, how it’s structured, and how messages flow between hosts, clients, and servers.
1
lightbulb
What Is MCP & Why It Exists
The N×M integration problem and how MCP is the “USB-C for AI.”
2
account_tree
Architecture: Hosts, Clients & Servers
The three-layer model, security isolation, and 1:N topology.
3
swap_horiz
Transports: stdio & Streamable HTTP
Local vs. remote communication, SSE, session management, and reconnection.
Primitives

Tools, Resources, Prompts & Sampling

The four core capabilities that MCP servers expose to AI applications.
4
build
Tools: Letting the Model Act
Tool discovery, JSON Schema, annotations, progress reporting, and change notifications.
5
folder_open
Resources & Resource Templates
URI-based data, text vs. binary, subscriptions, and embedded resources.
6
chat
Prompts: Reusable Workflows
Server-defined templates, multi-step workflows, and dynamic prompt generation.
7
psychology
Sampling & Elicitation
Server-initiated LLM calls, human-in-the-loop via elicitation, and the trust model.
Production

Security, Building & Ecosystem

OAuth, building servers, and deploying MCP in production with agent frameworks.
8
shield
Security, Authorization & OAuth
OAuth 2.1 + PKCE, dynamic client registration, roots, and consent scoping.
9
terminal
Building an MCP Server
Python & TypeScript SDKs, project structure, MCP Inspector, and IDE integration.
10
rocket_launch
MCP in Production & the Ecosystem
Registry, multi-server composition, agent frameworks, gateways, and the 2026 roadmap.