What Is MCP?
The Model Context Protocol (MCP) is an open standard that lets AI models connect to external tools and data sources. It’s what turns Claude from a smart assistant into a connected agent.
The Problem MCP Solves
Section titled “The Problem MCP Solves”Without MCP, Claude Code can:
- Read and write files on your machine
- Run shell commands
- Search your codebase
That’s powerful, but limited. What about:
- Your Notion workspace?
- Your Cloudflare dashboard?
- Your database on a remote server?
- Your project management tool?
- Your company’s internal APIs?
MCP bridges this gap. It gives Claude a standardized way to discover and use external tools.
How MCP Works
Section titled “How MCP Works”┌──────────────┐ ┌──────────────┐ ┌──────────────┐│ Claude Code │────▶│ MCP Server │────▶│ External ││ (Client) │◀────│ (Bridge) │◀────│ Service │└──────────────┘ └──────────────┘ └──────────────┘- MCP Client (Claude Code) discovers available tools from MCP servers
- MCP Server exposes tools that Claude can call (like “create_page”, “query_database”)
- External Service (Notion, Cloudflare, GitHub, etc.) processes the actual request
The MCP server acts as a translator between Claude’s tool-calling format and the external service’s API.
What MCP Servers Provide
Section titled “What MCP Servers Provide”An MCP server can expose three types of capabilities:
Actions Claude can take — creating records, querying data, sending messages.
Resources
Section titled “Resources”Data Claude can read — documents, database records, configuration.
Prompts
Section titled “Prompts”Pre-built prompt templates for common tasks with that service.
Available MCP Servers
Section titled “Available MCP Servers”The MCP ecosystem is growing fast. Some notable servers:
| Server | What It Connects To |
|---|---|
| Cloudflare | Workers, D1, KV, R2, Pages |
| Notion | Pages, databases, comments |
| GitHub | Issues, PRs, repos |
| Slack | Channels, messages |
| PostgreSQL | Direct database queries |
| Filesystem | Extended file operations |
| Brave Search | Web search results |
Why MCP Matters
Section titled “Why MCP Matters”Before MCP, connecting an AI to a new service required:
- Custom integration code
- Service-specific prompt engineering
- Brittle API wrappers
With MCP:
- One protocol to learn
- Servers are reusable across AI clients
- Community-maintained servers for popular services
- You can build your own for internal tools
MCP in This Course
Section titled “MCP in This Course”We’ll use MCP extensively:
- Module 3 (this module) — connecting to services and building servers
- Module 5 — Cloudflare MCP for cloud infrastructure
- Module 6 — MCP in automation pipelines
- Module 7 — building projects with MCP integrations