Database MCP Servers

Database access is the killer app for enterprise MCP. When employees can ask Claude "What were our top-selling products last quarter?" and get an instant, accurate answer from live data—that's transformative.

This chapter shows you how to build production-ready database MCP servers that are secure, performant, and enterprise-ready.

What You'll Learn

SectionTopics
The Enterprise Data Access ProblemWhy database access is MCP's killer app, the friction it eliminates
Building db-explorerStep-by-step server creation, query tools, schema introspection
SQL Safety and Injection PreventionSecurity patterns, parameterized queries, allowlisting
Resource-Based Data PatternsWhen to use resources vs tools, structured access patterns
Handling Large ResultsPagination, streaming, cursor-based navigation

Quick Preview

By the end of this chapter, you'll build a database server that lets Claude:

User: "Show me our top 10 customers by revenue"

Claude: I'll query the sales database for you.

[Calls list_tables tool]
[Calls query tool with: SELECT customer_name, SUM(order_total) as revenue 
 FROM orders GROUP BY customer_id ORDER BY revenue DESC LIMIT 10]

Here are your top 10 customers by revenue:
1. Acme Corp - $1,234,567
2. GlobalTech - $987,654
...

The Architecture

┌─────────────────────────────────────────────────────────┐
│                     Claude / AI Client                   │
└─────────────────────────┬───────────────────────────────┘
                          │ MCP Protocol
                          ▼
┌─────────────────────────────────────────────────────────┐
│                   Database MCP Server                    │
│  ┌─────────────┐  ┌─────────────┐  ┌─────────────┐     │
│  │ list_tables │  │    query    │  │  Resources  │     │
│  │    Tool     │  │    Tool     │  │ (optional)  │     │
│  └──────┬──────┘  └──────┬──────┘  └──────┬──────┘     │
│         │                │                │             │
│         └────────────────┼────────────────┘             │
│                          ▼                              │
│              ┌───────────────────────┐                  │
│              │   Connection Pool     │                  │
│              │   (sqlx + Arc)        │                  │
│              └───────────┬───────────┘                  │
└──────────────────────────┼──────────────────────────────┘
                           │
                           ▼
                    ┌──────────────┐
                    │   Database   │
                    │  (SQLite,    │
                    │  PostgreSQL, │
                    │  MySQL)      │
                    └──────────────┘

Prerequisites

Before starting this chapter, you should have:

Sample Database

We'll use the Chinook database—a sample database representing a digital media store with customers, invoices, tracks, and artists.

# Download the sample database
curl -L -o chinook.db https://github.com/lerocha/chinook-database/raw/master/ChinookDatabase/DataSources/Chinook_Sqlite.sqlite

Chapter Sections

1. The Enterprise Data Access Problem

Understand why database access is MCP's killer app for enterprises:

  • The current friction in getting data to AI
  • How MCP eliminates the copy-paste workflow
  • Security considerations for enterprise data

2. Building db-explorer

Build a complete database MCP server step-by-step:

  • Creating the server with cargo pmcp
  • Implementing list_tables and query tools
  • Testing with MCP Inspector and Claude

3. SQL Safety and Injection Prevention

Master security patterns for database access:

  • SQL injection attacks and prevention
  • Parameterized queries with sqlx
  • Allowlisting vs blocklisting approaches
  • Defense in depth strategies

4. Resource-Based Data Patterns

Learn when to use MCP resources instead of SQL tools:

  • Resources for structured, predictable access
  • Tools for flexible, ad-hoc queries
  • Hybrid approaches for different use cases

5. Handling Large Results

Handle enterprise-scale data volumes:

  • Why OFFSET pagination fails at scale
  • Cursor-based pagination patterns
  • Streaming for very large results
  • Memory-safe result handling

Hands-On Exercises

After completing the lessons, practice with these exercises:

Chapter 3 Exercises

  • Exercise 1: Database Query Basics - Build list_tables and execute_query tools
  • Exercise 2: SQL Injection Review - Find and fix security vulnerabilities
  • Exercise 3: Pagination Patterns - Implement cursor-based pagination

Security Checklist

Before deploying any database MCP server to production:

  • Only SELECT queries allowed (no mutations)
  • Parameterized queries for all user input
  • Row limits enforced on all queries
  • Sensitive columns filtered (SSN, passwords, PII)
  • Connection pooling configured
  • Query timeout set
  • Audit logging enabled
  • Authentication required (OAuth in production)

Knowledge Check

Test your understanding after completing the chapter:

Start with The Enterprise Data Access Problem →