multi-postgres

multi-postgres-mcp-server is a focused MCP server that lets you manage and query many PostgreSQL databases from a single context. It exposes your configured PostgreSQL connections in a read-only MCP-enabled endpoint so AI tools and clients can request data without risking modifications. The server supports hot-reload of its JSON configuration, meaning you can add, remove, or re-label databases on the fly and the server updates without a restart. This makes it ideal for AI-assisted analysis workflows where you want a safe, centralized read-only view over multiple databases. When integrated with MCP clients like Claude Code, Cursor, Windsurf, or Cline, you can issue structured queries and receive context-rich responses that reference the appropriate PostgreSQL databases behind the scenes. The setup emphasizes safety and simplicity, focusing on read-only access to prevent accidental data changes while still enabling powerful cross-database insights.

Prerequisites:

• Node.js (recommended) or a compatible runtime for the MCP server distribution you use, or access to the provided binary release for your OS.
• Internet access to download the MCP server package or binary.

Option A: Using npm/npx (Node.js environment)

1. Ensure Node.js and npm are installed on your machine. You can verify with: node -v npm -v
2. Install or run the MCP server via npx (no global install required): npx -y multi-postgres-mcp-server
3. The first run will guide you through initial configuration and where to place your hot-reload JSON config file.

Option B: Using the prebuilt binary (recommended for most users)

1. Download the appropriate binary for your OS from the official releases page.
2. Extract the archive and run the executable from your terminal or double-click the Windows/macOS/Linux launcher as described in the release notes.
3. Prepare your JSON configuration file and ensure the server can read it (path can be passed as a flag if supported by the binary).

Option C: Docker (alternative)

1. Pull an image (if provided by the project): docker pull jamaluddin9/multi-postgres-mcp-server
2. Run the container with a mounted config file: docker run -i -v /path/to/config.json:/app/config.json jamaluddin9/multi-postgres-mcp-server
3. Ensure the container exposes the MCP endpoint and that the config.json is valid JSON.

Configuration file guidance:

• The server reads a hot-reload JSON file describing databases with labels, hosts, ports, users, and credentials. See the example in the README for structure guidance. Place this file in a location accessible to the server and point the server to it if required by your launch method.

Tips and common issues:

• Use read-only connections in the PostgreSQL side when possible to minimize risk.
• Because the server supports hot-reload, you can update the JSON configuration on the fly; verify JSON validity after edits to avoid runtime errors.
• Ensure network access between the MCP server and all configured PostgreSQL databases (firewalls, VPNs, and DNS resolution can affect connectivity).
• If your AI tool experiences timeouts, consider increasing the query timeout on the MCP client or batching large requests.
• When moving between environments (dev, staging, prod), maintain separate config files to avoid cross-environment data exposure.
• If you see startup errors mentioning missing config paths, confirm the path you passed is correct and that the file has valid JSON.