Best MCP Servers for Databases:

Best MCP Servers for Databases: PostgreSQL, MySQL, MongoDB & More

Connecting your AI to a database is one of the most powerful things you can do with MCP. Instead of copying SQL results into a chat window, your AI reads the schema, writes queries, and analyzes results directly.

This guide covers every major database MCP server — how to install each one, example queries, and the security practices you need to follow.

Why Database MCP Servers Matter

Without MCP, interacting with your database through AI looks like this:

Run a query in your database client

Copy the results

Paste into Claude/ChatGPT

Ask your question

Get a suggested query back

Copy it, run it, repeat

With a database MCP server:

Ask your AI "What were our top 10 products by revenue last month?"

The AI reads your schema, writes the query, runs it (with your approval), and gives you the answer

That workflow change saves hours per week for anyone who works with data.

PostgreSQL MCP Servers

Official PostgreSQL MCP Server (by Anthropic)

The reference implementation. Simple, reliable, read-only by default.

Install:

json
{
  "mcpServers": {
    "postgres": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-postgres",
        "postgresql://user:password@localhost:5432/mydb"
      ]
    }
  }
}

What it exposes:

query — Run read-only SQL queries
Schema information as resources (tables, columns, types)

Key feature: Read-only by default. The server wraps queries in transactions and rolls them back, so your data stays safe. Pricing: Free, open source Best for: Getting started. If you just want to ask questions about your PostgreSQL data, this is all you need.

See our full PostgreSQL MCP Server review for more details.

Supabase MCP Server

Full Supabase integration — not just the database, but auth, storage, and edge functions too.

Install:

json
{
  "mcpServers": {
    "supabase": {
      "command": "npx",
      "args": ["-y", "supabase-mcp-server"],
      "env": {
        "SUPABASE_URL": "https://your-project.supabase.co",
        "SUPABASESERVICEKEY": "your-service-key"
      }
    }
  }
}

What it exposes: Database queries, table management, auth user management, storage operations, edge function management Pricing: Free server (Supabase has its own pricing tiers) Best for: Supabase users who want AI-assisted full-stack development

MySQL MCP Servers

MySQL MCP Server (Community)

Install:

json
{
  "mcpServers": {
    "mysql": {
      "command": "npx",
      "args": ["-y", "mysql-mcp-server"],
      "env": {
        "MYSQL_HOST": "localhost",
        "MYSQL_PORT": "3306",
        "MYSQL_USER": "root",
        "MYSQL_PASSWORD": "your-password",
        "MYSQL_DATABASE": "mydb"
      }
    }
  }
}

What it exposes: Schema inspection, query execution, table management Pricing: Free, open source Best for: MySQL/MariaDB users who need basic database AI integration

DBHub (Multi-Database)

DBHub is a universal MCP server that supports MySQL, MariaDB, PostgreSQL, and SQL Server through a single interface.

Install:

json
{
  "mcpServers": {
    "dbhub": {
      "command": "npx",
      "args": [
        "-y",
        "dbhub",
        "--dsn",
        "mysql://user:pass@localhost:3306/mydb"
      ]
    }
  }
}

Supported DSN formats:

PostgreSQL: postgresql://user:pass@host:5432/db
MySQL: mysql://user:pass@host:3306/db
SQL Server: sqlserver://user:pass@host:1433/db
MariaDB: mariadb://user:pass@host:3306/db

Pricing: Free, open source Best for: Teams that use multiple database types and want one MCP server for all of them

MongoDB MCP Server

Official MongoDB MCP Server

Built and maintained by the MongoDB team. Works with both local MongoDB and Atlas clusters.

Install:

json
{
  "mcpServers": {
    "mongodb": {
      "command": "npx",
      "args": [
        "-y",
        "mongodb-mcp-server",
        "--connectionString",
        "mongodb://localhost:27017"
      ]
    }
  }
}

For MongoDB Atlas:

json
{
  "mcpServers": {
    "mongodb-atlas": {
      "command": "npx",
      "args": [
        "-y",
        "mongodb-mcp-server",
        "--connectionString",
        "mongodb+srv://user:pass@cluster.mongodb.net/mydb"
      ]
    }
  }
}

What it exposes: Collection browsing, document queries, aggregation pipelines, index management Pricing: Free, open source Best for: MongoDB users at any scale. Atlas support means it works with production cloud databases too.

SQLite MCP Server

Official SQLite MCP Server (by Anthropic)

Lightweight and perfect for local databases, prototyping, or embedded data.

Install:

json
{
  "mcpServers": {
    "sqlite": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-sqlite",
        "/path/to/your/database.db"
      ]
    }
  }
}

What it exposes: Query execution, schema inspection, table creation, data analysis Pricing: Free, open source Best for: Local development, prototyping, analyzing data in SQLite files

Redis MCP Server

Install:

json
{
  "mcpServers": {
    "redis": {
      "command": "npx",
      "args": ["-y", "redis-mcp-server"],
      "env": {
        "REDIS_URL": "redis://localhost:6379"
      }
    }
  }
}

What it exposes: Key-value operations, hash operations, list operations, key scanning Pricing: Free, open source Best for: Debugging Redis caches, exploring key structures, managing session data

Enterprise & Cloud Database Servers

MindsDB MCP Server

MindsDB acts as a unified MCP gateway to multiple enterprise databases — Oracle, PostgreSQL, MongoDB, Snowflake, and more.

What it does: Connects AI models directly to your database ecosystem, whether cloud-based, on-premises, or hybrid. Pricing: Free tier available; enterprise plans for production use (check mindsdb.com for current pricing) Best for: Enterprise teams with complex, multi-database environments

Google Cloud Toolbox for Databases

A secure, configurable MCP interface for executing pre-defined queries against PostgreSQL, MySQL, SQL Server, Neo4j, and Spanner.

Key feature: YAML-based configuration lets you define approved queries, so the AI can only run operations you've pre-authorized. Pricing: Free server (standard Google Cloud charges apply) Best for: Teams that need strict control over what queries AI can run

ClickHouse MCP Server

Analytics-focused MCP server for ClickHouse columnar databases.

Best for: Data teams running analytical queries on large datasets

Example Queries You Can Run

Once your database MCP server is connected, here are practical things to ask:

Data analysis: > "What's the average order value by customer segment for the last 90 days?" Schema understanding: > "Show me how the users, orders, and products tables are related" Query debugging: > "This query is slow — can you look at the schema and suggest indexes?" Data quality: > "Are there any users with duplicate email addresses?" Report generation: > "Generate a summary of our monthly revenue trend for the last 12 months"

The AI reads your actual schema, so it writes accurate queries for your specific tables and columns — not generic examples.

Security Best Practices

Connecting AI to your database is powerful, but it comes with risks. Follow these rules:

1. Use Read-Only Credentials

Create a dedicated database user with SELECT-only permissions:

sql
-- PostgreSQL
CREATE USER mcpreadonly WITH PASSWORD 'securepassword';
GRANT SELECT ON ALL TABLES IN SCHEMA public TO mcp_readonly;
ALTER DEFAULT PRIVILEGES IN SCHEMA public GRANT SELECT ON TABLES TO mcp_readonly;

sql
-- MySQL
CREATE USER 'mcpreadonly'@'localhost' IDENTIFIED BY 'securepassword';
GRANT SELECT ON mydb.* TO 'mcp_readonly'@'localhost';

2. Never Use Production Credentials Directly

Connect to a read replica or a separate analytics database. Never point MCP at your primary production database with write access.

3. Restrict Table Access

Only grant access to tables the AI actually needs. Don't expose user passwords, payment info, or PII tables:

sql
-- Only grant access to specific tables
GRANT SELECT ON orders, products, categories TO mcp_readonly;
-- NOT: GRANT SELECT ON ALL TABLES

4. Review Every Query

MCP clients show you each query before it runs. Always review the SQL — especially check for:

Queries against sensitive tables

Large table scans without WHERE clauses

Unexpected JOINs that might expose data

5. Use Connection Limits

Set connection limits on your MCP database user to prevent runaway queries:

sql
-- PostgreSQL
ALTER USER mcp_readonly CONNECTION LIMIT 3;
ALTER USER mcpreadonly SET statementtimeout = '30s';

For a deeper dive, read our MCP security best practices guide.

Choosing the Right Database MCP Server

| Database | Best Server | Write Support | Maintained By |
|---|---|---|---|
| PostgreSQL | Official Postgres MCP | Read-only | Anthropic |
| MySQL/MariaDB | DBHub | Read + Write | Community |
| MongoDB | Official MongoDB MCP | Read + Write | MongoDB |
| SQLite | Official SQLite MCP | Read + Write | Anthropic |
| SQL Server | DBHub | Read + Write | Community |
| Redis | Redis MCP Server | Read + Write | Community |
| Multi-DB | MindsDB | Configurable | MindsDB |
| Supabase | Supabase MCP | Full access | Community |

What's Next

See our complete MCP servers guide for servers beyond databases
Set up MCP in your editor: Cursor guide | VS Code guide
Read the MCP security guide before connecting production databases

We track MCP support across 500+ AI tools in our directory — browse MCP-compatible tools to find the right fit for your database stack.

Best MCP Servers for Databases: PostgreSQL, MySQL, MongoDB & More