Exasol MCP Server

Provides an LLM access to the Exasol database via MCP tools. Includes the tools for reading the database metadata and executing data reading queries.

Features

• Collects the metadata.
  • Enumerates the existing database objects, including schemas, tables, views, functions and UDF scripts.
  • Provides a filtering mechanisms to use with object enumeration.
  • Describes the database objects: for tables returns the list of columns and constraints; for functions and scripts - the list of input and output parameters.
• Executes provided data reading SQL query. Disallows any other type of query.

Prerequisites

• Python >= 3.10.
• MCP Client application, e.g. Claude Desktop.

Installation

Ensure the uv package is installed. If uncertain call

uv --version

To install uv on macOS please use brew, i.e.

brew install uv

For other operating systems, please follow the instructions in the uv official documentation.

Using the server with the Claude Desktop.

To enable the Claude Desktop using the Exasol MCP server, the latter must be listed in the configuration file claude_desktop_config.json. A similar configuration file would exist for most other MCP Client applications.

To find the Claude Desktop configuration file, click on the Settings and navigate to the “Developer” tab. This section contains options for configuring MCP servers and other developer features. Click the “Edit Config” button to open the configuration file in the editor of your choice.

Add the Exasol MCP server to the list of MCP servers as shown in this configuration example.

{
  "mcpServers": {
    "exasol_db": {
      "command": "uvx",
      "args": ["exasol-mcp-server@latest"],
      "env": {
        "EXA_DSN": "my-dsn",
        "EXA_USER": "my-user-name",
        "EXA_PASSWORD": "my-password"
      }
    },
    "other_server": {}
  }
}

With these settings, uv will execute the latest version of the exasol-mcp-server in an ephemeral environment, without installing it.

Alternatively, the exasol-mcp-server can be installed using the command:

uv tool install exasol-mcp-server@latest

For further details on installing and upgrading the server using uv see the uv Tools documentation.

If the server is installed, the Claude configuration file should look like this:

{
  "mcpServers": {
    "exasol_db": {
      "command": "exasol-mcp-server",
      "env": "same as above"
    }
  }
}

Please note that any changes to the Claude configuration file will only take effect after restarting Claude Desktop.

Running modes

The MCP server can be deployed either locally, as described above, or as a remote HTTP server. To run the server as a Direct HTTP Server execute the command:

exasol-mcp-server-http --host <server-host> --port <server-port>

The host defaults to 0.0.0.0.

This command provides a simple way to verify the setup for a remote MCP Server deployment. For the production environment, one might consider using an ASGI server like Unicorn. The most flexible approach is implementing a wrapper for the Exasol MCP server that will provide the desired control options. For further information and ideas, please check the HTTP Deployment in the FastMCP documentation.

Here is an example code creating the Exasol MCP server from a wrapper.

from exasol.ai.mcp.server import mcp_server

exasol_mcp = mcp_server()

Configuration settings

The server is configured using environment variables and optionally a json file. In the above example, the server is provided with the database connection parameters, all other settings left to default. For the information on how to customize server settings please see the Server Setup User Guide.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Safe Harbor Statement: Exasol MCP Server & AI Solutions

Exasol’s AI solutions (including MCP Server) are designed to enable intelligent, autonomous, and highly performant access to data through AI and LLM-powered agents. While these technologies unlock powerful new capabilities, they also introduce potentially significant risks.

By granting AI agents access to your database, you acknowledge that the behavior of large language models (LLMs) and autonomous agents cannot be fully predicted or controlled. These systems may exhibit unintended or unsafe behavior—including but not limited to hallucinations, susceptibility to adversarial prompts, and the execution of unforeseen actions. Such behavior may result in data leakage, unauthorized data generation, or even data modification or deletion.

Exasol provides the tools to build AI-native workflows; however, you, as the implementer and system owner, assume full responsibility for managing these solutions within your environment. This includes establishing appropriate governance, authorization controls, sandboxing mechanisms, and operational guardrails to mitigate risks to your organization, your customers, and their data.