SQLGenius MCP Server

The T3Ta SQL MCP Server is a TypeScript‑based Model Context Protocol (MCP) server that bridges AI assistants and PostgreSQL databases. It solves the problem of securely exposing database queries to language models while keeping credentials and network topology hidden from the model. By running as a local or containerized process, it can be invoked via standard stdin/stdout streams, making it compatible with a wide range of AI tools such as Claude Desktop, Cursor IDE, and memory‑bank‑mcp-server.

At its core, the server offers a read‑only query engine powered by the library. Developers can send an MCP “call_tool” message containing a SQL string, and the server returns the results in JSON format. The read‑only nature guarantees that models cannot modify data, which is essential for compliance and security in production environments. When a database resides behind a private network or an AWS RDS instance, the server automatically establishes an SSH bastion tunnel if configured. This allows developers to keep their databases off‑premises while still giving the AI a direct, secure path for queries.

Key capabilities include:

• SSH Bastion Support – automatically spins up an SSH tunnel to a bastion host, routing traffic to the target PostgreSQL instance without exposing it publicly.
• Environment‑Driven Configuration – all connection parameters are read from environment variables or a file, enabling easy deployment in CI pipelines, Docker containers, or local machines.
• STDIO MCP Transport – the server listens on stdin and writes responses to stdout, making it a drop‑in replacement for any MCP client that communicates over the command line.
• Extensible Tool Interface – while currently focused on a single tool, the architecture allows additional tools (e.g., schema introspection) to be added with minimal friction.

Real‑world use cases abound: a data analyst can ask an AI assistant to “list the top 10 customers by revenue” and receive a JSON array without ever exposing credentials; a developer can prototype SQL queries in an IDE that talks to the MCP server, letting the AI suggest optimizations; or a DevOps engineer can run diagnostics on an RDS instance through a conversational interface, with the server ensuring all traffic is routed securely via SSH. The combination of secure tunneling, read‑only enforcement, and standard MCP compliance gives the T3Ta SQL MCP Server a distinct advantage in any environment where AI meets sensitive data.