Postgres MCP Pro

Add to your client

Copy the config for your MCP client and paste it into its config file.

{
  "mcpServers": {
    "postgres-mcp-pro": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "-e",
        "DATABASE_URI",
        "crystaldba/postgres-mcp",
        "--access-mode=restricted"
      ],
      "env": {
        "DATABASE_URI": "postgresql://<user>:<password>@<host>:5432/<dbname>"
      }
    }
  }
}

Before you start

• A reachable PostgreSQL database and a connection string in the form postgresql://user:password@host:5432/dbname
• Python 3.12+ with the uv or pipx package manager, OR Docker if you prefer the container image
• Recommended Postgres extensions for full functionality: pg_stat_statements (query stats) and hypopg (hypothetical indexes for tuning)
• An MCP-capable client (Claude Desktop, Cursor, Windsurf, etc.)

About Postgres MCP Pro

Tools & capabilities (9)

When to use it

• Use it when a query is slow and you want the agent to propose and validate the right index via EXPLAIN and hypothetical-index simulation
• Use it when you want a workload-wide index tuning pass instead of guessing at single indexes
• Use it when you need a database health checkup — bloat, cache hit ratio, vacuum status, replication lag, invalid constraints
• Use it when you want a read-only, resource-capped agent connection that is safe to point at production
• Use it when you want an agent to explore an unfamiliar schema (schemas, tables, constraints, indexes)
• Use it when you need ad-hoc SQL execution from your AI client during development

Quick setup

1. 1Install via pipx install postgres-mcp, uv pip install postgres-mcp, or docker pull crystaldba/postgres-mcp
2. 2(Recommended) Enable the pg_stat_statements and hypopg extensions in your database
3. 3Add the server to your MCP client config, setting DATABASE_URI to your connection string
4. 4Choose an access mode with --access-mode restricted (production) or unrestricted (development)
5. 5Restart the client, then verify by asking the agent to list schemas or run analyze_db_health

Postgres MCP Pro FAQ