Keboola MCP Server

Name: Keboola MCP Server
Author: keboola

@keboola

Repository

About Keboola MCP Server

Model Context Protocol (MCP) Server for the Keboola Platform

Basic information

Config

Add this server to your MCP-compatible client using the configuration below.

{
  "mcpServers": {
    "keboola-mcp-server": {
      "command": "uvx",
      "args": [
        "keboola_mcp_server",
        "--transport",
        "streamable-http"
      ]
    }
  }
}

Tools

No tools detected

We auto-extract tools from the README. The maintainer can list them under a ## Tools heading to populate this section.

Overview

What is Keboola MCP Server?

Keboola MCP Server is an open-source bridge that connects your Keboola project to AI agents and MCP clients such as Cursor, Claude, Windsurf, and VS Code. It exposes Keboola features—storage access, SQL transformations, job triggers, and more—as callable tools for AI assistants, eliminating the need for glue code.

How to use Keboola MCP Server?

The easiest way is to use the remotely hosted server: navigate to Keboola Project Settings → MCP Server, copy the server URL (e.g., https://mcp.<YOUR_REGION>.keboola.com/mcp), and paste it into your AI assistant’s MCP settings. For local setup, clone the repository, set environment variables (KBC_STORAGE_TOKEN, KBC_STORAGE_API_URL, KBC_WORKSPACE_SCHEMA, optionally KBC_BRANCH_ID), install dependencies with uv, and start the server with a transport option (stdio or streamable-http).

Key features of Keboola MCP Server

Query tables and manage table or bucket descriptions (Storage)
Create, list, and inspect extractors, writers, data apps, and transformation configurations (Components)
Create SQL transformations using natural language
Run components and transformations, retrieve job execution details (Jobs)
Build and manage workflow pipelines with Conditional and Orchestrator Flows
Create, deploy, and manage Streamlit Data Apps over storage data
Search, read, and update project documentation and metadata via natural language
Work safely in development branches isolated from production

Use cases of Keboola MCP Server

AI agents querying Keboola storage tables directly from a chatbot
Generating SQL transformations from natural language prompts
Automating job runs and monitoring execution from an AI assistant
Building and orchestrating complex data pipelines without manual configuration
Managing project metadata and documentation interactively

FAQ from Keboola MCP Server

What is the Remote MCP Server?

It is a hosted version of the server available on every multi-tenant Keboola stack. It supports OAuth authentication and eliminates the need for local setup.

How do I authenticate when using Keboola MCP Server?

For the remote server, authentication is done via OAuth: a browser window prompts you to log in with your Keboola account and select your project. For local setup, you provide a Storage API token (master or custom) via the KBC_STORAGE_TOKEN environment variable.

What are the requirements for local setup?

Python 3.10+ and uv are required. You also need access to a Keboola project with admin rights and a valid Storage API token. For custom tokens, you must additionally provide KBC_WORKSPACE_SCHEMA.

Can I restrict which tools are available to the AI agent?

Yes. When using HTTP transports, you can set HTTP headers such as X-Allowed-Tools, X-Disallowed-Tools, or X-Read-Only-Mode to control which tools the client can call.

How do development branches work?

You can scope all operations to a development branch by setting the KBC_BRANCH_ID environment variable or, on remote transports, by including the X-Branch-Id header in each request. If not set, the server uses the production branch by default.