mcp-postgres
Enables AI agents to interact with PostgreSQL databases through schema intelligence, query execution, and DBA tooling including index analysis and health monitoring. Features configurable access levels and audit logging for secure database operations.
README
mcp-postgres
MCP server for PostgreSQL. Gives AI agents schema intelligence, query execution, and DBA tooling — through the Model Context Protocol.
Unlike generic database MCP servers, mcp-postgres is Postgres-native. It extracts table/column comments, understands Postgres-specific catalog views, provides index analysis, and ships with configurable access levels so you don't hand an LLM unrestricted database access.
Features
Schema Intelligence
- List schemas, tables, views with sizes and row counts
- Full table descriptions: columns, types, constraints, indexes, foreign keys
- Extracts
COMMENT ONmetadata — gives the LLM semantic context about what columns mean - Search objects by name or comment across the entire database
Query Execution
- Read-only
querytool with automatic row limiting - Write-capable
executetool gated by access level EXPLAIN ANALYZEwith human-readable output
DBA Tooling
- Table stats: live/dead tuples, bloat percentage, vacuum history, scan patterns
- Index analysis: usage stats, unused index detection, missing index suggestions
- Database health: connections, cache hit ratio, long-running queries, throughput
Safety
- Four access levels:
readonly,readwrite,admin,unrestricted - SQL statement classification (SELECT, DML, DDL, admin) with enforcement
- Audit logging to stderr (JSON, one entry per query)
Quick Start
npx mcp-postgres --connection-string "postgres://user:pass@localhost:5432/mydb"
Or with environment variables:
DATABASE_URL="postgres://user:pass@localhost:5432/mydb" npx mcp-postgres
Claude Desktop
Add to your claude_desktop_config.json:
{
"mcpServers": {
"postgres": {
"command": "npx",
"args": [
"-y",
"mcp-postgres",
"--connection-string",
"postgres://user:pass@localhost:5432/mydb"
]
}
}
}
Claude Code
Add to your project's .mcp.json:
{
"mcpServers": {
"postgres": {
"command": "npx",
"args": ["-y", "mcp-postgres"],
"env": {
"DATABASE_URL": "postgres://user:pass@localhost:5432/mydb"
}
}
}
}
Tools
| Tool | Description | Access |
|---|---|---|
list_schemas |
List schemas with table counts and sizes | readonly |
list_tables |
List tables with comments, row counts, sizes | readonly |
describe_table |
Full table description with columns, indexes, FKs, comments | readonly |
search_objects |
Search objects by name or comment | readonly |
query |
Execute SELECT queries | readonly |
execute |
Execute INSERT/UPDATE/DELETE/CREATE/etc | varies |
explain_query |
EXPLAIN (ANALYZE) with readable output | readonly* |
table_stats |
Table statistics, bloat, vacuum info | readonly |
index_analysis |
Index usage, unused indexes, missing index hints | readonly |
database_health |
Connections, cache ratio, long queries, bloat | readonly |
*explain_query with analyze=true executes the query, so it respects the statement's access level.
Resources
| URI | Description |
|---|---|
postgres://schema/{name} |
Full DDL for a schema (CREATE TABLE statements with comments) |
postgres://extensions |
Installed PostgreSQL extensions |
Prompts
| Prompt | Description |
|---|---|
explore-database |
Guided database exploration — schemas, tables, relationships |
optimize-query |
Analyze a slow query with EXPLAIN, indexes, and recommendations |
health-check |
Comprehensive database health assessment |
Configuration
CLI Options
--connection-string PostgreSQL connection URL
--access-level readonly|readwrite|admin|unrestricted (default: readonly)
--row-limit Max rows returned per query (default: 500)
--schema Default schema filter (default: public)
--audit Enable query audit logging to stderr
Environment Variables
| Variable | Description |
|---|---|
DATABASE_URL |
PostgreSQL connection URL |
POSTGRES_URL |
Alternative connection URL |
MCP_POSTGRES_ACCESS_LEVEL |
Access level override |
MCP_POSTGRES_ROW_LIMIT |
Row limit override |
Access Levels
| Level | SELECT | INSERT/UPDATE/DELETE | CREATE/ALTER/DROP | TRUNCATE/DROP DATABASE |
|---|---|---|---|---|
readonly |
yes | no | no | no |
readwrite |
yes | yes | no | no |
admin |
yes | yes | yes | no |
unrestricted |
yes | yes | yes | yes |
Default is readonly. Use the minimum level needed.
Audit Logging
Enable with --audit. Logs every tool invocation to stderr as JSON:
{"timestamp":"2026-04-03T12:00:00.000Z","tool":"query","sql":"SELECT * FROM users","statementType":"select","accessLevel":"readonly","allowed":true,"durationMs":12,"rowCount":42}
Pipe stderr to a file to capture: mcp-postgres --audit 2>audit.log
Architecture
src/
├── index.ts Entry point and CLI
├── server.ts MCP server setup
├── config.ts Configuration parsing
├── db/
│ ├── pool.ts Connection pool management
│ └── query.ts Query execution with timing
├── tools/
│ ├── schema.ts Schema exploration tools
│ ├── query.ts Query execution tools
│ └── performance.ts DBA and health tools
├── resources/
│ └── schema.ts Schema DDL resources
├── prompts/
│ └── index.ts Prompt templates
└── safety/
├── classifier.ts SQL statement classification
├── access.ts Access level enforcement
└── audit.ts Audit logging
Development
npm install
npm test # run tests
npm run build # compile TypeScript
npm run dev -- --connection-string "postgres://..." # run in dev mode
License
MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.