MCP Servers — Connecting External Tools

Claude’s Plugin System

Up to this point, Claude Code has been working with what’s in front of it — your local files, your terminal, your git repo. That’s powerful, but it’s also a walled garden. What if you want Claude to check your GitHub issues? Comment on a PR? Query a database? Post to Slack?

That’s what MCP does. The Model Context Protocol is how Claude connects to the outside world. Think of MCP servers as plugins — each one gives Claude a new set of abilities. The GitHub MCP server lets Claude read issues, create PRs, and comment on code reviews. The filesystem MCP server lets Claude access directories outside the current project. The Slack MCP server lets Claude post messages. And there are dozens more.

The beauty of MCP is that Claude treats these external tools exactly like its built-in tools. It doesn’t “call an API” or “run a script” — it just has new capabilities that show up in its toolbox, as natural as reading a file or running a command.

What We’ll Cover

1. What MCP is — the Model Context Protocol and why it exists
2. How MCP servers work — processes that expose tools to Claude
3. Configuration — setting up .mcp.json
4. GitHub MCP in practice — issues, PRs, and comments
5. Filesystem MCP — accessing directories outside your project
6. Permissions — controlling which MCP tools Claude can use
7. The MCP ecosystem — where to find more servers

Why This Block Matters

Here’s the real power move: when Claude can interact with both your code and your project management tools, it stops being just a coding assistant and becomes a development workflow assistant.

Imagine this: “Look at the open issues labeled bug, find the one about the dark theme toggle, read the relevant source files, implement a fix, create a branch, push it, and open a PR linking back to the issue.” That’s not science fiction — that’s a Claude session with MCP configured.

MCP is what turns Claude from “a really smart pair programmer” into “an engineer who can navigate your entire development ecosystem.”

Prerequisites

• Completed Blocks 0-8
• A GitHub account with access to a repository (ai-coderrank or any test repo)
• The gh CLI installed and authenticated (gh auth status should show you logged in)
• Node.js installed (MCP servers run as Node processes)

Choose Your Format

Pick the format that matches how you are using the block: