mcp -unitycatalog
Unity Catalog AI Model Context Protocol Server
claude mcp add --transport stdio ognis1205-mcp-server-unitycatalog uvx run mcp-server-unitycatalog --uc_server <your Unity Catalog server URL> --uc_catalog <your catalog name> --uc_schema <your schema name>
How to use
This MCP server exposes Unity Catalog functions as MCP tools. In addition to the Unity Catalog Functions registered in your Unity Catalog instance, you can access a set of predefined MCP tools such as uc_list_functions, uc_get_function, uc_create_function, and uc_delete_function. These tools let you list available functions, fetch details for a specific function, create a new Python-based function (note: this API is experimental), and delete existing functions within a parent catalog and schema. The server is designed to work with the Unity Catalog API, so you provide the Unity Catalog server URL, catalog, and schema, and optionally supply an access token and verbose logging level. Typical usage involves running the MCP server using uv (the Universal Verifier) or via a Docker image, then invoking the tools through your MCP client of choice.
How to install
Prerequisites:
- Install UV (and optionally UVX) per your environment documentation: https://docs.astral.sh/uv/
- Access to a Unity Catalog instance with a catalog and schema to target
- Network access from your environment to the Unity Catalog server
Installation steps:
- Install UV and UVX (if you haven’t already).
- macOS/Linux: curl -fsSL https://uv.sh/get | bash
- Windows: follow the official UV installation instructions
- Clone or obtain the mcp-server-unitycatalog repository locally.
- Install any runtime dependencies if needed (usually handled by UV when launching the server).
- Run the MCP server using UV with the proper configuration options (see the example below).
Additional notes
Environment variables:
- UC_SERVER: Unity Catalog base URL
- UC_CATALOG: Catalog name
- UC_SCHEMA: Schema name
- UC_TOKEN (optional): Access token for Unity Catalog API
- UC_VERBOSITY (optional): Logging level (default: warn)
- UC_LOG_DIRECTORY (optional): Directory for log files (default: .mcp_server_unitycatalog) Common issues:
- Ensure network access from the MCP host to the Unity Catalog server is allowed.
- If using Docker, make sure the image mcp/unitycatalog is accessible and the container can access your Unity Catalog endpoint.
- If uc_create_function is used, provide a valid Python script containing the function to register.
Configuration options can be overridden via CLI, environment variables, or .env files as supported by the UV-based launcher.
Related MCP Servers
web-agent-protocol
🌐Web Agent Protocol (WAP) - Record and replay user interactions in the browser with MCP support
mcp-reddit
A Model Context Protocol (MCP) server that provides tools for fetching and analyzing Reddit content.
edumcp
EDUMCP is a protocol that integrates the Model Context Protocol (MCP) with applications in the education field, dedicated to achieving seamless interconnection and interoperability among different AI models, educational applications, smart hardware, and teaching AGENTs.
mkinf
mkinf SDK to interact with mkinf hub MCP servers
mcp-tool-kit
Agentic abstraction layer for building high precision vertical AI agents written in python for Model Context Protocol.
fhir
FHIR MCP Server – helping you expose any FHIR Server or API as a MCP Server.