MCP Iceberg Catalog

The MCP Iceberg Catalog is a Model Context Protocol server that bridges Claude Desktop with an Apache Iceberg data lake. By exposing a lightweight SQL interface, it allows AI assistants to query, describe, and manipulate Iceberg tables directly from the Claude UI, eliminating the need for separate command‑line tools or custom connectors. This solves a common pain point for data scientists and ML engineers: accessing large, partitioned datasets in a lakehouse while still enjoying the conversational workflow of an AI assistant.

At its core, the server implements three intertwined components. First, the MCP protocol handler manages communication over stdio, translating Claude’s JSON messages into executable actions and returning results in the same format. Second, a query processor parses standard SQL statements—such as , , , and —using a robust parser library, ensuring that the assistant can express complex queries in natural language. Finally, the Iceberg integration layer leverages PyIceberg and PyArrow to perform catalog operations, schema management, and efficient data scans, all while keeping the response latency low enough for interactive use.

Key capabilities include dynamic table discovery (), schema introspection (), ad‑hoc querying, and data ingestion via . These features empower developers to prototype ETL pipelines, validate data quality, or generate analytics reports without leaving the Claude environment. The server also supports future extensions such as , , and table creation, making it a flexible foundation for evolving lakehouse workflows.

Real‑world scenarios that benefit from this MCP server include data exploration in machine learning projects, rapid prototyping of feature engineering pipelines, and operational monitoring of data lake health. For instance, a data engineer can ask Claude to “show me the latest 10 rows from the sales table” and receive an immediate tabular response, all while the assistant can suggest further transformations or visualizations. Because the server communicates via MCP, it seamlessly integrates with other AI tools that already consume or produce structured data, enabling end‑to‑end pipelines from natural language to actionable insights.

What sets the MCP Iceberg Catalog apart is its focus on developer productivity and AI‑centric workflows. By abstracting away the complexities of Iceberg’s REST catalog, authentication, and S3 storage details, it lets teams prototype faster and iterate on data models in a conversational manner. The modular architecture also makes it straightforward to add new SQL operations, support additional data types, or plug in advanced security and monitoring features as the project scales.