Step 3 — MCP — concepts and clients

MCP (Model Context Protocol) is the standard that connects AI assistants with external systems. Anthropic open-sourced it in 2024, and it became the de-facto standard fast.

Why

Before MCP:

Each tool needed bespoke plugins (Cursor and Claude Code didn't share)
Adding a tool = a lot of code

After MCP:

One server works everywhere (Claude Code, Cursor, Zed, Cline…)
Three primitives: Tools, Resources, Prompts

The three primitives

Primitive	What	Example
Tools	callable functions	`slack_send_message`, `github_create_pr`
Resources	readable data	files, DB rows, API responses
Prompts	predefined commands	"review this code"

A minimal server (Python)

from mcp.server.fastmcp import FastMCP

mcp = FastMCP("My First MCP")

@mcp.tool()
def add(a: int, b: int) -> int:
    """add two integers"""
    return a + b

@mcp.resource("config://app/version")
def get_version() -> str:
    return "1.0.0"

if __name__ == "__main__":
    mcp.run()

Tools (add) + Resources (config://app/version) in one file.

Wire into Claude Code

claude mcp add my-server -- python my-mcp-server.py

Now claude can call add(3, 5).

Public servers worth knowing

Server	Use
`@modelcontextprotocol/server-filesystem`	local files
`@modelcontextprotocol/server-github`	GitHub PRs / issues
`@modelcontextprotocol/server-slack`	Slack
`@upstash/context7-mcp`	live docs
`@executeautomation/playwright-mcp-server`	browser automation

claude mcp list shows registered servers.

Try it

Save the snippet, register, then ask "add 5 and 7" inside claude.

Going deeper

Step 4 — Skills, Subagents, Hooks.

Why