PostgreSQL MCP Server
Enables secure, read-only PostgreSQL database interaction through natural language, with automatic database discovery and connection management.
README
PostgreSQL MCP Server
A secure, read-only PostgreSQL Model Context Protocol (MCP) server for AI assistant integration with automatic database discovery and connection management.
What It Does
This tool provides two main components:
- Query Executor (
execute_query.py): Interactive PostgreSQL query execution with support for direct queries, file input, and interactive mode - MCP Server (
mcp_postgresql_server.py): AI assistant integration that allows natural language database interactions through Claude and other MCP clients
Key Features
- Automatic Database Discovery: Scans project files for database configurations and presents options for selection
- Read-Only Safety: Blocks write operations by default (configurable)
- Interactive Configuration: Guides users through database setup
with automatic
.envfile management - Connection Pooling: Efficient PostgreSQL connection management
- AI Integration: Works seamlessly with Claude Desktop and Cursor IDE
Quick Start
Prerequisites: Ensure you have uv installed.
Using uvx (Recommended)
Run directly without installation:
# MCP Server for AI integration
uvx mcp-postgresql-server
# Query executor
uvx --from mcp-postgresql-server execute-query "SELECT version()"
Configuration
Set your database connection:
export MCP_POSTGRESQL_DATABASE="postgres://username:password@hostname:port/database"
Or create a .env file:
MCP_POSTGRESQL_DATABASE=postgres://username:password@hostname:port/database
MCP_POSTGRESQL_READ_ONLY=true
MCP_POSTGRESQL_LOG_FILE=./mcp-postgresql.log
MCP_POSTGRESQL_LOG_LEVEL=info
Usage Examples
Query Executor
Using uvx (recommended):
# Interactive mode
uvx --from mcp-postgresql-server execute-query
# Direct query
uvx --from mcp-postgresql-server execute-query "SELECT COUNT(*) FROM users"
# From file
uvx --from mcp-postgresql-server execute-query --file queries.sql
Using Python directly:
# Interactive mode
python3 execute_query.py
# Direct query
python3 execute_query.py "SELECT COUNT(*) FROM users"
# From file
python3 execute_query.py --file queries.sql
MCP Server with Claude Code (globally)
Add to ~/.claude.json:
{
"mcpServers": {
"mcp-postgres": {
"command": "uvx",
"args": ["mcp-postgresql-server"],
"env": {
"MCP_POSTGRESQL_CWD": "${PWD}"
}
}
}
}
Then ask Claude:
- "Show me all tables in the database"
- "Execute SELECT COUNT(*) FROM users"
- "Help me write a query to find recent orders"
Cursor IDE Integration
Create/update mcp.json in your project root:
{
"mcpServers": {
"mcp-postgresql-server": {
"command": "uvx",
"args": ["mcp-postgresql-server"],
"env": {
"MCP_POSTGRESQL_CWD": "."
}
}
}
}
Configuration Discovery
The server automatically discovers database configurations from:
.envfiles.confand.inifiles.jsonand.yamlfiles- Individual parameter files (db.host, db.user, etc.)
When multiple configurations are found, it presents an interactive selection menu.
Saved configurations are stored in a .env file for future use
using MCP_POSTGRESQL_DATABASE variable.
Environment Variables
| Variable | Description | Default |
|---|---|---|
MCP_POSTGRESQL_DATABASE |
PostgreSQL connection URI | Required |
MCP_POSTGRESQL_READ_ONLY |
Enable read-only mode | true |
MCP_POSTGRESQL_LOG_FILE |
Log file path (optional) | None |
MCP_POSTGRESQL_LOG_LEVEL |
Log level (debug, info, warning, error, critical) | error |
MCP_POSTGRESQL_CWD |
Path of the project working directory | . |
Security
- Read-only by default: Blocks INSERT, UPDATE, CREATE, ALTER, DROP operations
- Connection validation: Validates all database connections before use
- Credential protection: Supports environment variables and
.envfiles - Error handling: Comprehensive error reporting without exposing sensitive data
Common Issues
Connection errors: Verify MCP_POSTGRESQL_DATABASE format: postgres://user:password@host:port/database
Permission issues: Ensure database user has appropriate SELECT permissions
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。