Greenplum MCP Server: Configuring Access to Multiple Databases

Products

• VMware Tanzu Data Suite
• VMware Tanzu Data Suite

Issue/Introduction

Users or administrators may want to know if a single instance of the VMware Tanzu Greenplum MCP (Model Context Protocol) server can simultaneously connect to and serve multiple databases located within the same Greenplum cluster.

This article addresses whether the VMware Tanzu Greenplum Model Context Protocol (MCP) server can be configured to connect to multiple databases simultaneously within a single Greenplum cluster and provides the recommended method for cross-database operations.

Environment

VMware Tanzu Greenplum MCP Server
Version: 18.4.0

Cause

By design, the Greenplum MCP server initializes a connection pool at startup targeting a single specific database defined in its configuration. This ensures secure, governed access and optimized performance for AI-driven natural language queries.

Resolution

Can I configure multiple databases in the config.yaml?

No. The Greenplum MCP server is designed to establish a connection pool to a single database at a time. Therefore, you cannot list multiple target databases within the static configuration file (config.yaml) for a single server instance.

How to interact with multiple databases

While the server binds to one database context, you can direct the AI client to perform operations on different databases within the same cluster using Target Parameters and Contextual Instructions.

1. Client-Side Instructions

You can pass the target database as a parameter in your client window. Instruct the AI client to route calls by providing a prompt such as:

"Execute all subsequent MCP tool calls on the target database 'finance'."

This ensures the AI client identifies the specific database context for subsequent tool executions.

2. Permissions Requirements

For the above method to work, the credentials (username/password) defined in your MCP server config.yaml or environment variables must have the appropriate permissions to access the target database specified in the tool call.

3. Multiple Server Instances (Alternative)

If your use case requires strict isolation or different credential sets for different databases, the recommended architecture is to deploy multiple instances of the MCP server (e.g., in separate Docker containers), each configured with a unique GPDB_DATABASE value.