Natural Language to Kusto Query MCP Server

What is Natural Language to Kusto Query MCP Server?

It is a configurable MCP server that converts natural language prompts into Kusto queries and optionally executes those queries against a Kusto database. It is built for developers and data analysts who work with Kusto databases and want to query them using plain English.

How to use Natural Language to Kusto Query MCP Server?

Configure a settings.yaml file with your Azure OpenAI model and Kusto database details. Run the server via Docker or dotnet run and invoke the tools list-supported-tables, generate-kusto-query, or execute-kusto-query with the required parameters (table, category, prompt, and optionally outputType).

Key features of Natural Language to Kusto Query MCP Server

• Converts natural language into Kusto queries
• Generates queries without executing them
• Generates and executes queries, outputting JSON or CSV
• Supports multiple Kusto databases via a single config file
• Accepts environment variable references in configuration

Use cases of Natural Language to Kusto Query MCP Server

• Query a Kusto database without knowing KQL syntax
• Automate data exploration across multiple Kusto databases
• Generate Kusto query drafts for review before execution
• Integrate natural-language querying into VS Code workflows

FAQ from Natural Language to Kusto Query MCP Server

What dependencies are required?

Azure OpenAI endpoint and deployment, a Kusto database endpoint, and either Docker or the .NET runtime. Optionally, an Azure OpenAI key and a Kusto access token can be provided.

How is authentication handled?

Authentication uses an Azure OpenAI key (or default Azure credentials) for the model and a Kusto access token (or default Azure credentials) for the database. Default credentials fallback only works when running with dotnet run inside VS Code.

Where do the queried data live?

All data resides in the Kusto databases configured in the settings.yaml file. The server itself does not store any data.

What transport does the server use?

The server is configured to use the stdio transport, as shown in the VS Code MCP configuration example.

Are there any known limits?

The fallback to default Azure credentials for both the model and Kusto database works only when running the server with dotnet run inside VS Code, not with Docker.