🔌 MCP — Connecting Claude to the World

The Model Context Protocol (MCP) lets you connect Claude Code to external tools, databases, and services — giving it superpowers beyond your local files.

What is MCP?

How MCP works

An MCP server is a small program that runs on your computer (or a remote server) and exposes tools that Claude can call. For example:

• A database MCP server exposes tools like query_database(sql) and list_tables()
• A GitHub MCP server exposes tools like create_pr(title, body) and list_issues()
• A Zotero MCP server exposes tools like search_library(query) and get_citation(key)

You configure which servers to run in ~/.claude/settings.json, and Claude automatically has access to their tools.

Adding MCP servers

You add MCP servers to your settings file:

// ~/.claude/settings.json
{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/Users/you/research"]
    },
    "github": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "env": {
        "GITHUB_PERSONAL_ACCESS_TOKEN": "your_token_here"
      }
    }
  }
}

Or add them via the command line (easier for beginners):

# Add the official Anthropic filesystem server
claude mcp add filesystem npx @modelcontextprotocol/server-filesystem ~/research

# Add a GitHub server
claude mcp add github npx @modelcontextprotocol/server-github

# List all configured servers
claude mcp list

MCP servers for researchers

Research example: Query a database directly

# With SQLite MCP server configured:
> Connect to my database at data/survey.db and:
  1. Show me all table names
  2. Count observations in the main household table
  3. Find all villages where more than 50% of households took up insurance
  4. Export that list to a CSV file

Research example: Search your Zotero library

> Search my Zotero library for papers about
  "index insurance smallholder Africa" and give me:
  - The 5 most relevant papers with their key findings
  - A properly formatted bibliography entry for each
  - Which ones I should cite in my literature review

Building a simple MCP server

You can build your own MCP server to connect Claude to any tool in your research stack. Here's a minimal example in Python:

# A simple MCP server that reads Stata datasets
from mcp.server import Server
from mcp.server.models import InitializationOptions
import mcp.types as types

app = Server("stata-reader")

@app.list_tools()
async def handle_list_tools() -> list[types.Tool]:
    return [
        types.Tool(
            name="read_stata",
            description="Read a Stata .dta file and return summary info",
            inputSchema={
                "type": "object",
                "properties": {
                    "filepath": {"type": "string"}
                }
            }
        )
    ]

@app.call_tool()
async def handle_call_tool(name: str, arguments: dict):
    if name == "read_stata":
        import pandas as pd
        df = pd.read_stata(arguments["filepath"])
        return [types.TextContent(
            type="text",
            text=str(df.describe())
        )]

Module 12 (Agents & API) goes deeper into building custom MCP servers.