MCP Server

The Exasol MCP (Model Context Protocol) Server enables you to interact with your Exasol database using plain English - no SQL knowledge required. It integrates seamlessly with AI assistants like Claude Desktop, allowing you to ask questions about your data in natural language and get results instantly.

What is the MCP Server?

The MCP Server acts as a bridge between Large Language Models (LLMs) and your Exasol database. It provides AI assistants with tools to:

  • Explore your database schema - Discover tables, views, functions, and their structures

  • Search for database objects - Find tables and columns using keywords

  • Execute queries - Run SQL queries generated from natural language requests

  • Understand your data - Get column definitions, constraints, and relationships

Instead of writing SQL like this:

SELECT customer_name, SUM(order_total)
FROM orders
WHERE order_date >= '2024-01-01'
GROUP BY customer_name
ORDER BY SUM(order_total) DESC
LIMIT 10;

You can simply ask:

“Show me the top 10 customers by total order value this year”

The MCP Server translates your request into SQL and returns the results.

🚀 Key Features

Metadata Discovery

Enumerate all database objects including schemas, tables, views, functions, and UDF scripts. Filter and search to find exactly what you need.

Object Descriptions

Get detailed information about tables (columns, constraints) and functions (input/output parameters).

Keyword Search

Find database objects by searching for keywords across names and descriptions.

Query Execution

Run SQL queries generated by the AI assistant based on your natural language requests.

Prerequisites

Before getting started, ensure you have:

  • Python 3.10 or higher

  • An MCP Client application (e.g., Claude Desktop)

  • The uv package manager

Installing uv

Check if uv is already installed:

uv --version

On macOS, install via Homebrew:

brew install uv

For other operating systems, follow the uv installation guide.

Quick Setup with Claude Desktop

The easiest way to get started is using the Exasol MCP Server with Claude Desktop.

Step 1: Find the Configuration File

  1. Open Claude Desktop

  2. Click on Settings

  3. Navigate to the Developer tab

  4. Click Edit Config to open claude_desktop_config.json

Step 2: Add the Exasol MCP Server

Add the following configuration to enable the MCP server:

{
    "mcpServers": {
        "exasol_db": {
            "command": "uvx",
            "args": ["exasol-mcp-server@latest"],
            "env": {
                "EXA_DSN": "your-host:8563",
                "EXA_USER": "your-username",
                "EXA_PASSWORD": "your-password"
            }
        }
    }
}

Replace the connection parameters with your actual Exasol database credentials:

  • EXA_DSN: Your database host and port (e.g., mydb.exasol.com:8563)

  • EXA_USER: Your database username

  • EXA_PASSWORD: Your database password

Note

With this configuration, uv will run the latest version of the MCP server in an ephemeral environment without permanently installing it.

Step 3: Restart Claude Desktop

Changes to the configuration file only take effect after restarting Claude Desktop. Close and reopen the application.

Step 4: Start Querying!

Once configured, you can start asking questions about your data in Claude:

  • “What tables are available in the SALES schema?”

  • “Show me the structure of the CUSTOMERS table”

  • “List all orders from the last 30 days”

  • “What’s the average order value by region?”

The AI assistant will use the MCP Server to explore your database, generate appropriate SQL queries, and return the results in a conversational format.

Installing the MCP Server Permanently

If you prefer to install the MCP server instead of running it ephemerally, use:

uv tool install exasol-mcp-server@latest

Then update your Claude configuration to use the installed version:

{
    "mcpServers": {
        "exasol_db": {
            "command": "exasol-mcp-server",
            "env": {
                "EXA_DSN": "your-host:8563",
                "EXA_USER": "your-username",
                "EXA_PASSWORD": "your-password"
            }
        }
    }
}

To upgrade to the latest version:

uv tool upgrade exasol-mcp-server

Running as a Remote HTTP Server

For production deployments or team-wide access, you can run the MCP server as an HTTP server.

Basic HTTP Server

Start the server with:

exasol-mcp-server-http --host 0.0.0.0 --port 8000

This creates a simple HTTP endpoint that AI clients can connect to remotely.

Production Deployment

For production environments, consider using an ASGI server like Uvicorn or implementing a custom wrapper. Here’s an example of creating a custom MCP server instance:

from exasol.ai.mcp.server import mcp_server

# Create the MCP server instance
exasol_mcp = mcp_server()

For more deployment options and best practices, see the FastMCP HTTP Deployment Guide.

Practical Examples

Here are some examples of natural language queries you can use with the MCP Server:

Schema Exploration

  • “What databases/schemas exist?”

  • “List all tables in the SALES schema”

  • “Show me all views related to customer data”

Table Structure

  • “Describe the ORDERS table”

  • “What are the columns in the CUSTOMER_TRANSACTIONS table?”

  • “Show me the constraints on the PRODUCTS table”

Data Analysis

  • “How many orders were placed last month?”

  • “What’s the total revenue by product category?”

  • “Show me customers who haven’t placed an order in 90 days”

  • “What are the top 5 best-selling products?”

Function Discovery

  • “What UDF functions are available?”

  • “Show me the parameters for the CALCULATE_DISCOUNT function”

Advanced Configuration

The MCP server supports additional configuration options through environment variables and JSON configuration files.

For detailed information on customizing server settings, authentication options (including OpenID), BucketFS access, and other advanced features, see the Server Setup Guide in the official documentation.

Using with Other AI Assistants

While this guide focuses on Claude Desktop, the Exasol MCP Server works with any MCP-compatible client. The configuration pattern is similar:

  1. Install or run the MCP server

  2. Configure your AI client with the connection details

  3. Provide Exasol database credentials via environment variables

Consult your AI assistant’s documentation for specific MCP server configuration instructions.

Security Considerations

Important

Safe Harbor Statement

The behavior of Large Language Models (LLMs) and autonomous AI agents cannot be fully predicted or controlled. These systems may exhibit unintended behavior including:

  • Hallucinations or incorrect queries

  • Susceptibility to adversarial prompts

  • Execution of unforeseen actions

Such behavior may result in data leakage, unauthorized data access, or unintended data modifications.

Best Practices:

  • Use read-only database users when possible to prevent unintended data modifications

  • Implement proper access controls - limit the MCP server to specific schemas or tables

  • Monitor query execution - review what queries are being generated and executed

  • Sandbox sensitive environments - don’t connect production databases directly without proper safeguards

  • Review security settings regularly and audit AI assistant interactions with your database

Learn More

Feedback

Last updated: February 2026