mcpField
P
PostgreSQL MCP Server

PostgreSQL MCP Server

Local
github.com/musaddiq-devDatabases1Updated 23w ago

About PostgreSQL MCP Server

A Python MCP server for inspecting PostgreSQL schemas and querying databases. It combines read-only analysis tools with a separately available write tool that can modify data or schema.

Key features

  1. 01Discover tables and schemas
  2. 02Run bounded read-only SQL queries
  3. 03Explain query plans
  4. 04Inspect relationships and indexes
  5. 05Preview rows and SQL routines
  6. 06Access table metadata resources

Use cases

  1. 01Explore an unfamiliar database schema
  2. 02Review query plans and index details
  3. 03Retrieve bounded data for analysis

MCP support

Exposes PostgreSQL inspection and query tools, along with table-list and schema resources, to MCP-compatible clients. Read queries run in read-only transactions; a separate write tool can modify data or schema when permitted by the database user.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

mcp.json · Universal config
{
"mcpServers": {
"musaddiq-dev-postgresql": {
"command": "uvx",
"args": [
"mdev-postgresql-mcp-server"
],
"env": {
"POSTGRES_HOST": "<POSTGRES_HOST>",
"POSTGRES_PORT": "<POSTGRES_PORT>",
"POSTGRES_USER": "<POSTGRES_USER>",
"POSTGRES_PASSWORD": "<POSTGRES_PASSWORD>",
"POSTGRES_DB": "<POSTGRES_DB>",
"POSTGRES_READ_QUERY_LIMIT": "<POSTGRES_READ_QUERY_LIMIT>"
}
}
}
}
Run commanduvx mdev-postgresql-mcp-server
Environment variables to set
POSTGRES_HOSTPostgreSQL host
POSTGRES_PORTPostgreSQL port
POSTGRES_USERPostgreSQL username
POSTGRES_PASSWORDSecretPostgreSQL password
POSTGRES_DBPostgreSQL database name
POSTGRES_READ_QUERY_LIMITMaximum rows returned by read queries

Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.

Packages
pypimdev-postgresql-mcp-server0.1.2

Tools

No tool list yet

Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.

Details

From the project README

PostgreSQL MCP Server

A Python Model Context Protocol (MCP) server for inspecting and querying PostgreSQL databases from MCP-compatible clients. It provides schema discovery, safe read-only query execution, query explanation, table previews, index analysis, relationship inspection, and PostgreSQL resources for table metadata.

Features

  • List public tables and inspect table schemas
  • Execute read-only SQL in a PostgreSQL read-only transaction
  • Explain query plans without executing the target query directly
  • Preview table rows with a fixed limit
  • Inspect foreign-key relationships and indexes
  • Expose passive MCP resources for table lists and schema details

Safety Model

postgresql_execute_read_query runs with PostgreSQL read-only transaction mode, caps returned rows by POSTGRES_READ_QUERY_LIMIT, and rolls back after execution. The server also includes postgresql_execute_write_query, which only accepts a single INSERT, UPDATE, DELETE, CREATE, ALTER, DROP, or TRUNCATE statement and can modify data/schema if the connected database user has permission. Do not auto-approve write-capable tools in your MCP client. For public or shared use, run the server with a dedicated read-only PostgreSQL user.

Requirements

  • Python 3.11+
  • PostgreSQL database
  • MCP-compatible client such as Claude Desktop, Cursor, VS Code, or another MCP host

Installation

When published to PyPI, install or run the server like a standard Python MCP package:

uvx mdev-postgresql-mcp-server

For local development from source:

git clone https://github.com/musaddiq-dev/postgresql-mcp-server.git
cd postgresql-mcp-server
python -m venv .venv
source .venv/bin/activate
pip install -e .

Configuration

Copy the example environment file and update it with your database connection details.

cp .env.example .env
VariableDescriptionRequiredDefault
POSTGRES_HOSTPostgreSQL hostYeslocalhost
POSTGRES_PORTPostgreSQL portYes5432
POSTGRES_USERPostgreSQL usernameYesNone
POSTGRES_PASSWORDPostgreSQL passwordNoNone
POSTGRES_DBPostgreSQL database nameYesNone
LOG_LEVELPython logging level written to stderrNoINFO
POSTGRES_READ_QUERY_LIMITMaximum rows returned by read queriesNo1000

Example read-only user:

CREATE USER mcp_readonly WITH PASSWORD 'change-me';
GRANT CONNECT ON DATABASE your_database TO mcp_readonly;
GRANT USAGE ON SCHEMA public TO mcp_readonly;
GRANT SELECT ON ALL TABLES IN SCHEMA public TO mcp_readonly;
ALTER DEFAULT PRIVILEGES IN SCHEMA public GRANT SELECT ON TABLES TO mcp_readonly;

Running

mdev-postgresql-mcp-server

From a local checkout before PyPI publication, run:

python -m postgresql_mcp_server.server

MCP Client Configuration

For published installs, prefer uvx. MCP servers using stdio must write protocol messages only to stdout; this server writes logs to stderr through Python logging.

Claude Desktop / Cursor / Windsurf / Cline

Most MCP clients accept this mcpServers JSON shape:

{
  "mcpServers": {
    "postgresql": {
      "command": "uvx",
      "args": ["mdev-postgresql-mcp-server"],
      "env": {
        "POSTGRES_HOST": "localhost",
        "POSTGRES_PORT": "5432",
        "POSTGRES_USER": "mcp_readonly",
        "POSTGRES_PASSWORD": "change-me",
        "POSTGRES_DB": "your_database"
      }
    }
  }
}

For local development from this repository, use the installed console script path instead:

{
  "mcpServers": {
    "postgresql": {
      "command": "/absolute/path/to/postgresql-mcp-server/.venv/bin/mdev-postgresql-mcp-server",
      "args": [],
      "env": {
        "POSTGRES_HOST": "localhost",
        "POSTGRES_PORT": "5432",
        "POSTGRES_USER": "mcp_readonly",
        "POSTGRES_PASSWORD": "change-me",
        "POSTGRES_DB": "your_database"
      }
    }
  }
}
Claude Code CLI
claude mcp add postgresql \
  --env POSTGRES_HOST=localhost \
  --env POSTGRES_PORT=5432 \
  --env POSTGRES_USER=mcp_readonly \
  --env POSTGRES_PASSWORD=change-me \
  --env POSTGRES_DB=your_database \
  -- uvx mdev-postgresql-mcp-server
VS Code MCP

VS Code uses the same command/args/env model in its MCP configuration:

{
  "servers": {
    "postgresql": {
      "type": "stdio",
      "command": "uvx",
      "args": ["mdev-postgresql-mcp-server"],
      "env": {
        "POSTGRES_HOST": "localhost",
        "POSTGRES_PORT": "5432",
        "POSTGRES_USER": "mcp_readonly",
        "POSTGRES_PASSWORD": "change-me",
        "POSTGRES_DB": "your_database"
      }
    }
  }
}

Tools

ToolPurposeSafety
postgresql_list_tablesList public base tablesRead-only
postgresql_describe_tableShow columns and metadata for a tableRead-only
postgresql_execute_read_queryRun bounded SQL under read-only transaction modeRead-only
postgresql_execute_write_queryRun a single approved modifying SQL statement and commitDestructive
postgresql_explain_queryReturn PostgreSQL EXPLAIN output for a single queryRead-only
postgresql_get_database_summaryReturn database version and table countRead-only
postgresql_get_relationshipsInspect foreign-key relationshipsRead-only
postgresql_analyze_indexesInspect indexes and sizesRead-only
postgresql_preview_tableReturn up to 10 rows from a tableRead-only
postgresql_search_sql_definitionsSearch public SQL routines/functionsRead-only

Resources

  • postgres://list_tables returns public table names.
  • postgres://schema/{table_name} returns a generated schema statement for a table.

Smoke Check

Without a database, verify syntax with:

python -m py_compile src/postgresql_mcp_server/server.py

With a configured database, start the server and use your MCP client to call list_tables.

Distribution

This server is published through the standard Python MCP distribution path:

  • PyPI package: mdev-postgresql-mcp-server
  • MCP Registry name: io.github.musaddiq-dev/postgresql-mcp-server
  • Runtime hint: uvx
  • Transport: stdio

The mcp-name marker at the top of this README is required for MCP Registry ownership verification. Users should prefer uvx mdev-postgresql-mcp-server in local MCP client configurations.

Security Notes

  • Do not commit .env or MCP client configs containing credentials.
  • Use least-privilege database users.
  • Treat execute_write_query as destructive and require explicit user approval in your MCP client.
  • Review generated SQL before running write-capable tools.

License

MIT

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.