Keboola MCP Server

Keboola’s MCP Server is a lightweight, open‑source bridge that exposes the full breadth of Keboola’s data platform to modern AI assistants. By turning storage, transformations, job triggers, and workflow orchestration into MCP‑compatible tools, it eliminates the need for custom adapters or glue code. Developers can now query tables, run SQL transformations, and trigger pipelines directly from Claude, Cursor, LangChain, Amazon Q, or any other MCP‑enabled client—all without leaving the AI environment.

The core problem it solves is the friction between data platform and conversational AI. Traditional approaches require writing SDK wrappers, managing authentication tokens, or exposing APIs behind custom endpoints. Keboola’s MCP Server abstracts these complexities: a single SSE endpoint, OAuth authentication, and branch‑scoped requests are all that is needed. The server automatically translates natural‑language commands into Keboola API calls, returning results in a format that the assistant can display or act upon.

Key capabilities include:

• Storage access – Query tables, list buckets, and edit metadata through intuitive prompts.
• Component management – Create, list, and inspect extractors, writers, data apps, and transformation configurations.
• SQL transformations – Generate SQL jobs from plain English, then deploy them as Keboola transformations.
• Job orchestration – Launch components or flows, monitor execution status, and retrieve logs.
• Flows – Build Conditional and Orchestrator flows to automate complex pipelines, all controllable via the assistant.
• Data Apps – Deploy Streamlit‑based dashboards that surface query results directly within the AI chat.
• Metadata operations – Search, read, and update project documentation or object metadata using natural language.
• Branch safety – Work in Keboola development branches with the header, ensuring that experiments do not touch production data.

In real‑world scenarios, a data analyst can ask Claude to “run a sales forecast transformation” and receive the updated table instantly. A data engineer might trigger an orchestrator flow to refresh nightly ETL jobs, while a product manager can deploy a new data app and share it with the team—all through conversational commands. The server’s tight integration with OAuth means that permissions are respected automatically, giving teams granular control over who can execute which operations.

Overall, Keboola’s MCP Server offers a unified, secure, and developer‑friendly interface that brings the power of Keboola’s data ecosystem into the next generation of AI workflows.