Ainative Postgres MCP server

ainative-postgres-mcp

Zero-config PostgreSQL MCP server with auto-provisioning. Drop-in replacement for @modelcontextprotocol/server-postgres — no DATABASE_URL needed.

The official @modelcontextprotocol/server-postgres requires you to bring your own Postgres database and pass a connection string. This fork auto-provisions a managed PostgreSQL instance with pgvector on first run. Agents get a database instantly with zero configuration.

Quick Start

npx ainative-postgres-mcp

That's it. On first run:

1. A free managed PostgreSQL instance is provisioned
2. pgvector extension is enabled automatically
3. Connection details are saved to ~/.ainative/postgres-config.json, .env, and .mcp.json
4. A claim URL is printed so you can take ownership of the database

MCP Configuration

Claude Code / Claude Desktop

Add to your claudedesktopconfig.json or .mcp.json:

{
  "mcpServers": {
    "postgres": {
      "command": "npx",
      "args": ["-y", "ainative-postgres-mcp"]
    }
  }
}

With existing database

{
  "mcpServers": {
    "postgres": {
      "command": "npx",
      "args": ["-y", "ainative-postgres-mcp"],
      "env": {
        "DATABASE_URL": "postgresql://user:pass@host:5432/mydb"
      }
    }
  }
}

Cursor

Add to .cursor/mcp.json:

{
  "mcpServers": {
    "postgres": {
      "command": "npx",
      "args": ["-y", "ainative-postgres-mcp"]
    }
  }
}

Windsurf

Add to ~/.windsurf/mcp.json:

{
  "mcpServers": {
    "postgres": {
      "command": "npx",
      "args": ["-y", "ainative-postgres-mcp"]
    }
  }
}

Tools

Comparison with official server-postgres

How Auto-Provisioning Works

Start
  |
  v
DATABASE_URL set? ----yes----> Connect directly
  |
  no
  v
~/.ainative/postgres-config.json exists? ----yes----> Load and connect
  |
  no
  v
.mcp.json has DATABASE_URL? ----yes----> Load and connect
  |
  no
  v
POST /api/v1/public/instant-db   (get API key)
  |
  v
POST /api/v1/zerodb/postgres/provision   (get Postgres)
  |
  v
Save to ~/.ainative/postgres-config.json + .env + .mcp.json
  |
  v
Connect and enable pgvector

pgvector Support

pgvector is automatically enabled on provisioned instances. You can use vector columns in your tables:

CREATE TABLE documents (
  id SERIAL PRIMARY KEY,
  content TEXT,
  embedding vector(1536)
);

And run similarity searches:

SELECT content, embedding <=> '[0.1, 0.2, ...]'::vector AS distance
FROM documents
ORDER BY distance
LIMIT 10;

Environment Variables

Free Tier

Shortened. The full README is on GitHub.