mcp-bauplan
@marcoeg
An MCP server to interact with data and run pipeline using Bauplan
Overview
What is mcp-bauplan?
mcp-bauplan is a minimalist Model Context Protocol (MCP) server that enables AI assistants to interact with data tables and run queries on Bauplan, a data store of Iceberg tables in S3.
How to use mcp-bauplan?
Install via uvx mcp-bauplan and configure it in your Claude Desktop config file with environment variables BAUPLAN_API_KEY, BAUPLAN_BRANCH, and BAUPLAN_NAMESPACE. Optionally set BAUPLAN_TIMEOUT. It supports both STDIO and SSE transports. For SSE, run uvx mcp-bauplan --transport sse --port 9090 with a .env file.
Key features of mcp-bauplan
- Lists all tables in the configured namespace
- Retrieves the schema of any data table
- Runs SELECT-only SQL queries on tables
- Supports both STDIO and SSE transports
- Configurable query timeout (default 30 seconds)
Use cases of mcp-bauplan
- Discover available Iceberg tables in a Bauplan namespace
- Inspect table schemas before writing queries
- Run ad‑hoc SELECT queries from an AI chat interface
- Validate data or explore dataset content without a separate client
FAQ from mcp-bauplan
What are the runtime dependencies?
Python 3.10 or higher, the uv package manager, and a valid Bauplan API key.
What transports are supported?
Both STDIO and SSE transports are supported. You can switch between them in the MCP Inspector during development.
How do I get a Bauplan API key?
Request one at https://www.bauplanlabs.com/#join.
What types of queries can I run?
Only SELECT queries are allowed. No data modification commands are supported.
Where are my data tables stored?
Bauplan manages Iceberg tables in S3. The server reads from the configured namespace and branch.
