PostgreSQL MCP Server
A production-ready MCP server that enables safe, read-only SQL SELECT queries against PostgreSQL databases with built-in security validation. It features connection pooling, automatic row limits, and structured logging to ensure secure and reliable database interactions.
README
PostgreSQL MCP Server
A production-ready Model Context Protocol (MCP) server that enables Claude to execute read-only SQL queries against PostgreSQL databases safely and securely.
Features
- Read-Only Queries: Only SELECT statements allowed - no data modification
- Security First: Comprehensive SQL injection prevention and query validation
- Connection Pooling: Efficient database connection management
- Row Limits: Automatic enforcement of query result limits
- Production Ready: Structured logging, error handling, and graceful shutdown
- Type Safe: Built with TypeScript for reliability
- VPS Ready: Includes deployment guides for remote VPS setups
Read the Deployment Guide for instructions on setting up the server on a VPS.
Installation
npm install
Configuration
Create a .env file based on .env.example:
# Required
DATABASE_URL=postgresql://user:password@localhost:5432/dbname
# Optional (with defaults)
NODE_ENV=development
LOG_LEVEL=info
DEFAULT_QUERY_LIMIT=100
MAX_QUERY_LIMIT=10000
QUERY_TIMEOUT_MS=30000
Environment Variables
DATABASE_URL(required): PostgreSQL connection stringNODE_ENV: Environment mode (development,production,test)LOG_LEVEL: Logging level (trace,debug,info,warn,error,fatal)DEFAULT_QUERY_LIMIT: Default row limit for queries (default: 100)MAX_QUERY_LIMIT: Maximum allowed row limit (default: 10000)QUERY_TIMEOUT_MS: Query execution timeout in milliseconds (default: 30000)
Development
# Run in development mode with auto-reload (no build needed)
npm run dev
# Build the project (compile TypeScript to JavaScript)
npm run build
# Run the production build (after npm run build)
npm start
# Run tests
npm test
# Run tests in watch mode
npm run test:watch
# Run tests with coverage
npm run test:coverage
# Type check
npm run type-check
# Lint
npm run lint
# Format code
npm run format
Local Testing
Test your MCP server locally before integrating with Claude Desktop:
Option 1: Automated Test Script (Recommended)
# Ensure .env file is configured with DATABASE_URL
npm run test:local
This runs automated tests that verify:
- Server starts correctly
- Tools are registered
- Queries execute successfully
- Security validation works
- Row limits are enforced
Option 2: MCP Inspector (Interactive)
# Build first
npm run build
# Start the inspector
npm run inspect
This opens a web UI where you can:
- Browse available tools
- Execute queries interactively
- View responses and debug messages
- Test different parameters
Option 3: Manual Testing with PostgreSQL Client
# Run the dev server
npm run dev
# In another terminal, test your database connection
psql $DATABASE_URL -c "SELECT 1"
Usage with Claude Desktop
1. Build the Project
npm run build
2. Configure Claude Desktop
Add to your Claude Desktop config file:
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"postgres": {
"command": "node",
"args": ["/absolute/path/to/mcp-server/dist/index.js"],
"env": {
"DATABASE_URL": "postgresql://user:password@localhost:5432/dbname"
}
}
}
}
3. Restart Claude Desktop
Restart Claude Desktop to load the MCP server.
4. Test the Connection
Try asking Claude:
- "List all tables in the database"
- "Show me the first 5 rows from the users table"
- "Count the total number of records in the orders table"
Available Tools
execute_query
Execute a read-only SQL SELECT query against the PostgreSQL database.
Parameters:
query(string, required): The SQL SELECT query to executelimit(number, optional): Row limit for results (default: from config, max: 10000)
Example:
{
"query": "SELECT id, name, email FROM users WHERE active = true",
"limit": 50
}
Response:
{
"success": true,
"data": {
"rows": [...],
"rowCount": 42,
"fields": [
{ "name": "id", "dataType": "integer" },
{ "name": "name", "dataType": "text" },
{ "name": "email", "dataType": "varchar" }
],
"executionTimeMs": 15
}
}
Security
Query Validation
The server implements multiple layers of security:
- SQL Parsing: Queries are parsed using
node-sql-parserto ensure valid syntax - Statement Type Check: Only SELECT statements are allowed
- Dangerous Function Detection: Blocks PostgreSQL functions like
pg_read_file,COPY, etc. - Comment Removal: SQL comments are stripped to prevent comment-based injection
- Row Limit Enforcement: Automatic LIMIT clauses prevent excessive data retrieval
- Read-Only Validation: Additional layer ensures no data modification
Blocked Operations
- INSERT, UPDATE, DELETE, TRUNCATE
- CREATE, DROP, ALTER (DDL operations)
- GRANT, REVOKE (permission changes)
- SELECT INTO (data copying)
- FOR UPDATE/FOR SHARE (row locking)
- Dangerous functions (pg_read_file, pg_ls_dir, COPY, etc.)
Best Practices
- Use Read-Only Database Users: Create a database user with SELECT-only permissions
- Limit Database Access: Only grant access to necessary schemas/tables
- Use SSL/TLS: Enable SSL for database connections in production
- Monitor Queries: Review logs regularly for suspicious activity
- Set Resource Limits: Configure appropriate query timeouts and row limits
Testing
Unit Tests
npm test
Unit tests cover:
- Query validation and sanitization
- Security checks (SQL injection, dangerous functions)
- Row limit enforcement
Integration Tests
Integration tests require a running PostgreSQL instance:
# Start PostgreSQL (example with Docker)
docker run --name test-postgres \
-e POSTGRES_PASSWORD=testpass \
-p 5432:5432 \
-d postgres:16
# Set DATABASE_URL and run tests
export DATABASE_URL=postgresql://postgres:testpass@localhost:5432/postgres
npm test
Architecture
src/
├── index.ts # Entry point
├── server.ts # MCP server setup
├── config/
│ └── index.ts # Configuration loader
├── database/
│ └── connection.ts # PostgreSQL connection with pooling
├── security/
│ ├── query-validator.ts # SQL validation
│ └── read-only-validator.ts # Read-only enforcement
├── tools/
│ └── query-tool.ts # execute_query tool
└── utils/
├── logger.ts # Structured logging
└── types.ts # TypeScript types
Error Handling
The server provides detailed error messages:
- ValidationError: Invalid query syntax or parameters
- SecurityError: SQL injection attempts or unauthorized operations
- DatabaseError: Connection failures or query execution errors
- ConfigurationError: Missing or invalid configuration
Logging
Structured JSON logging in production, pretty-printed in development:
{
"level": "info",
"time": "2024-01-11T10:30:00.000Z",
"msg": "Query executed successfully",
"rowCount": 42,
"executionTimeMs": 15
}
Troubleshooting
Connection Issues
- Verify
DATABASE_URLis correct - Check database is running and accessible
- Ensure firewall allows PostgreSQL connections
- Test connection with
psqlcommand
Permission Errors
- Ensure database user has SELECT permissions
- Check schema access permissions
- Verify connection string includes correct database name
Query Timeouts
- Increase
QUERY_TIMEOUT_MSfor long-running queries - Optimize slow queries with indexes
- Reduce row limits if fetching too much data
Future Enhancements (v2)
- Multi-database support
- Additional tools (list_tables, describe_table, get_schema)
- HTTP transport for remote access
- Schema caching
- Query history logging
- Prometheus metrics export
License
MIT
Contributing
Contributions are welcome! Please ensure:
- Tests pass (
npm test) - Code is formatted (
npm run format) - Types are valid (
npm run type-check) - Security best practices are followed
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。