PostgreSQL MCP Server for Claude Desktop
Enables Claude Desktop to interact with PostgreSQL databases through natural language for schema exploration, data analysis, and query execution. Users can search schemas, describe tables, and perform read or write operations without needing to write manual SQL.
README
🐘 PostgreSQL MCP Server for Claude Desktop
Connect Claude Desktop to your PostgreSQL database using the Model Context Protocol (MCP). Ask Claude questions in plain English and it will query your database, explore schemas, and analyze data — no SQL required.
✨ Features
| Category | Tools |
|---|---|
| 🗄️ Database Info | get_database_info, list_databases |
| 📂 Schema Exploration | list_schemas, list_tables, describe_table, search_schema, list_indexes |
| 🔍 Querying | execute_query, get_table_sample, get_table_stats, explain_query |
| ✏️ Write Operations | execute_write (disabled by default — opt-in via .env) |
| 📄 Resources | schema://overview, table://{schema}/{table} |
| 💬 Prompts | analyze_table, write_sql_query |
🚀 Quick Start
1. Clone the Repository
git clone https://github.com/sarotechhub/Claude-Desktop-to-PostgreSQL.git
cd Cluade-MCP-PostgreSQL
2. Create a Virtual Environment
# Windows
python -m venv .venv
.venv\Scripts\activate
# macOS / Linux
python -m venv .venv
source .venv/bin/activate
3. Install Dependencies
pip install -r requirements.txt
4. Configure Your Database
# Copy the example env file
cp .env.example .env
Edit .env with your PostgreSQL credentials:
DB_HOST=localhost
DB_PORT=5432
DB_NAME=your_database_name
DB_USER=your_username
DB_PASSWORD=your_password
# Set to "true" to allow INSERT/UPDATE/DELETE
ALLOW_WRITE_OPERATIONS=false
All credentials live only in
.env— they are never put in the Claude Desktop config file.
5. Test the Server
python main.py
You should see:
🚀 Starting PostgreSQL MCP Server...
✅ Database connection pool ready.
Press Ctrl+C to stop.
6. Connect to Claude Desktop
Find your Claude Desktop config file:
| OS | Path |
|---|---|
| Windows | %APPDATA%\Claude\claude_desktop_config.json |
| macOS | ~/Library/Application Support/Claude/claude_desktop_config.json |
| Linux | ~/.config/Claude/claude_desktop_config.json |
Add the postgres block to the mcpServers section (update the path):
{
"mcpServers": {
"postgres": {
"command": "[PATH_TO_YOUR_VENV_PYTHON_EXE]",
"args": ["[PATH_TO_YOUR_PROJECT_ROOT]\\main.py"],
"cwd": "[PATH_TO_YOUR_PROJECT_ROOT]",
"env": {
"PYTHONPATH": "[PATH_TO_YOUR_PROJECT_ROOT]"
}
}
}
}
Note: No DB credentials go in this file. The server reads them automatically from your
.envfile usingpython-dotenv.
Restart Claude Desktop after saving the config.
💬 Example Conversations with Claude
Once connected, try asking Claude:
"List all tables in my database"
"Describe the structure of the users table"
"Show me 10 sample rows from the orders table"
"How many rows are in each table in the public schema?"
"Find all columns related to 'email' across all tables"
"What indexes exist on the products table?"
"Write a query to find the top 10 customers by total order value"
"Explain why this query might be slow: SELECT * FROM orders WHERE status = 'pending'"
🔧 Tool Reference
Database Info
| Tool | Description |
|---|---|
get_database_info |
PostgreSQL version, DB size, connection count, server info |
list_databases |
All databases on the server with sizes and encoding |
Schema Exploration
| Tool | Description |
|---|---|
list_schemas() |
All user-defined schemas in the connected database |
list_tables(schema) |
Tables in a schema with row counts and sizes |
describe_table(table, schema) |
Columns, types, nullable, defaults, PKs, FKs |
search_schema(keyword) |
Find tables/columns by keyword (case-insensitive) |
list_indexes(table, schema) |
Indexes on a table with type and columns |
Querying
| Tool | Description |
|---|---|
execute_query(sql, limit) |
Read-only SELECT (enforced via read-only transaction) |
get_table_sample(table, schema, limit) |
Sample rows from a table (max 100) |
get_table_stats(table, schema) |
Row counts, sizes, vacuum/analyze timestamps |
explain_query(sql) |
EXPLAIN execution plan (no data modification) |
Write Operations
| Tool | Description |
|---|---|
execute_write(sql, confirm) |
INSERT/UPDATE/DELETE — requires ALLOW_WRITE_OPERATIONS=true in .env AND confirm=True |
🔒 Security
- Read-only by default — write operations require explicit opt-in in
.env - Credentials in
.envonly — never in the Claude Desktop config - Read-only transactions — SELECT queries run inside
readonly=Truetransactions - Input validation — table/schema names validated with regex to prevent injection
- No DDL — DROP, CREATE, ALTER are always blocked even when writes are enabled
.envexcluded from git — credentials never committed
📁 Project Structure
Cluade-MCP-PostgreSQL/
├── main.py # 🚀 MCP server entry point (FastMCP + stdio)
├── database.py # 🔌 Async connection pool (asyncpg)
├── tools/
│ ├── __init__.py
│ ├── schema_tools.py # list_schemas, list_tables, describe_table, search_schema, list_indexes
│ ├── query_tools.py # execute_query, get_table_sample, get_table_stats, explain_query
│ ├── write_tools.py # execute_write (opt-in)
│ └── database_tools.py # get_database_info, list_databases
├── .env # ✅ Your credentials (NOT committed to git)
├── .env.example # Template for new users
├── .gitignore # Excludes .env, .venv, __pycache__
├── requirements.txt # mcp[cli], asyncpg, python-dotenv, pydantic, orjson
├── claude_desktop_config.json # Example Claude Desktop config snippet
└── README.md
⚙️ Environment Variables
All configuration is done via .env:
| Variable | Default | Description |
|---|---|---|
DB_HOST |
localhost |
PostgreSQL host |
DB_PORT |
5432 |
PostgreSQL port |
DB_NAME |
— | Database name |
DB_USER |
— | Database user |
DB_PASSWORD |
— | Database password |
DB_MIN_CONNECTIONS |
1 |
Min pool connections |
DB_MAX_CONNECTIONS |
10 |
Max pool connections |
ALLOW_WRITE_OPERATIONS |
false |
Enable INSERT/UPDATE/DELETE |
LOG_LEVEL |
INFO |
Logging level (DEBUG/INFO/WARNING) |
🛠️ Troubleshooting
Claude doesn't see the MCP server
- Fully quit and reopen Claude Desktop after editing the config
- Check the
cwdpath — it must point to the project folder - Verify the
.venvPython path is correct:.venv\Scripts\python.exe
Connection refused / authentication failed
- Check your
.envcredentials match your PostgreSQL setup - Test directly:
psql -h localhost -U your_user -d your_db - Ensure PostgreSQL is running:
pg_isready
mcp or asyncpg module not found
.venv\Scripts\pip install -r requirements.txt
Server starts but tools don't appear in Claude
- Open Claude Desktop → Settings → Developer → MCP Servers
- Check for error messages next to the
postgresserver entry - Run
python main.pymanually and check stderr for import errors
📋 Requirements
- Python 3.10+
- PostgreSQL 12+
- Claude Desktop (with MCP support)
📄 License
MIT License — free to use, modify, and distribute.
🤝 Contributing
Pull requests welcome! Please fork the repo, create a feature branch, and submit a PR with a clear description.
Claude-Desktop-to-PostgreSQL
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。