Zaturn

What kdqed/zaturn MCP server does

The kdqed/zaturn MCP server connects AI-assisted analysis to a selection of database systems and local data files. It is designed for conversational exploration: users can ask questions about their data while the AI uses the available SQL execution capabilities to retrieve and analyze results.

Supported database sources include PostgreSQL, SQLite, DuckDB, MySQL, ClickHouse, SQL Server, and BigQuery. File-based inputs currently include CSV and Parquet. This makes the server suitable for workflows that combine warehouse data with local extracts or analytical files, provided those sources are supported by the installation.

The output is not limited to raw query results. Zaturn can provide tabular and textual summaries, and it can create image visualizations for supported chart types.

How it works

Zaturn can run as an MCP application or through a web interface described as similar to a Jupyter Notebook. In an MCP workflow, an AI model uses Zaturn's tools to run SQL against connected sources. The user can therefore describe an analysis goal in natural language rather than manually writing SQL or Python.

A typical interaction would involve connecting one or more supported sources, asking a question about the data, and reviewing the returned table, written interpretation, or chart. The README does not specify the individual MCP tool names, connection configuration fields, or the exact request flow for each source; those details are provided through the project's installation documentation.

Tools and capabilities

The kdqed/zaturn MCP server supports:

• SQL execution for PostgreSQL, SQLite, DuckDB, MySQL, ClickHouse, SQL Server, and BigQuery data sources.
• Analysis of CSV and Parquet files.
• Tabular result presentation.
• Textual summaries of analyzed data.
• Scatter and line plots.
• Histograms.
• Strip and box plots.
• Bar plots.
• Density heatmaps, also described as two-dimensional histograms.
• Polar scatter and polar line plots.

These capabilities fit exploratory analysis, business-question answering, and basic visual investigation. They do not establish that the project provides predictive modeling, machine learning, dashboards, or multi-user access; those items appear in the roadmap rather than the current feature list.

Limitations and notes

The currently documented source list is limited to the seven SQL systems plus CSV and Parquet files. Additional data sources are planned in response to community requests, so an unsupported system may require a project issue or future release.

The same applies to visualizations: the README identifies a defined set of available chart types and invites requests for more. Dashboard building, multi-user functionality, and predictive analytics or machine learning are listed as roadmap items, not current capabilities.

Installation instructions are maintained at https://zaturn.pro/docs/install. The provided material does not state a package-manager command, required environment variables, authentication method, client-specific setup, or software license. For those details, consult the project's installation documentation and repository configuration before deployment. The kdqed/zaturn MCP server is therefore best evaluated as a data analysis and visualization layer whose exact connection setup depends on the sources and deployment mode you choose.