mcp-server

Overview

The Keboola MCP Server is an open-source implementation of the Model Context Protocol that connects AI agents to the Keboola data platform. It transforms Keboola features—such as storage access, SQL transformations, and job triggers—into callable tools for AI assistants like Claude, Cursor, CrewAI, and LangChain. The server supports multiple transport protocols including stdio and HTTP, with both remote and local deployment options. It provides robust authentication through OAuth and granular tool authorization controls, making it suitable for production environments.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• query_data

  Query tables directly from Keboola storage

  Try:Show me sales data from the last quarterTry:Find all customers in the Netherlands
• create_config

  Create new configuration for extractors, writers, or data apps

  Try:Create a new CSV extractor for customer dataTry:Set up a transformation job
• get_configs

  List and inspect existing component configurations

  Try:Show me all extractors in the projectTry:List all transformation jobs
• run_job

  Execute components and transformations

  Try:Run the data refresh jobTry:Execute the sales transformation
• get_job_details

  Retrieve details and status of job executions

  Try:Check the status of the last jobTry:Show me the logs from the failed job
• create_sql_transformation

  Create SQL transformations using natural language

  Try:Create a transformation that calculates monthly revenueTry:Build a query to find top 10 products
• get_buckets

  List and manage storage buckets

  Try:Show me all buckets in the projectTry:List files in the marketing bucket
• get_tables

  Retrieve information about tables in storage

  Try:List all tables in the databaseTry:Show me columns of the customers table
• create_data_app

  Create and deploy Streamlit data applications

  Try:Build a data app for sales dashboardTry:Create an app to visualize customer data
• search_metadata

  Search project documentation and object metadata

  Try:Find documentation about the ETL processTry:Search for tables related to sales
• get_flow

  Retrieve details of workflow pipelines

  Try:Show me the daily sales pipelineTry:Check the status of the monthly reporting flow
• update_table_description

  Update descriptions for tables or buckets

  Try:Add a description to the sales tableTry:Update the documentation for the customers bucket

Note: Tool names inferred from feature descriptions and authorization documentation without explicit tool signatures

Comparable tools

Installation

{
  "mcpServers": {
    "keboola": {
      "command": "http",
      "args": ["https://mcp.<YOUR_REGION>.keboola.com/mcp"]
    }
  }
}

FAQ

What authentication methods are supported?

The server supports OAuth authentication for remote deployments and environment variables or headers for local setups.

Can I restrict which tools are available to AI agents?

Yes, you can use HTTP headers like X-Allowed-Tools and X-Disallowed-Tools to filter tool access, or set X-Read-Only-Mode to restrict to read-only operations.

Compare mcp-server with