MCP Datahub, developed by Txn2, is a free MCP server designed to connect AI assistants to DataHub metadata catalogs (v1.3 or later). Built with Go, it provides structured access to enterprise data for AI agents and developers. This tool enables AI clients like Claude and OpenAI to interact with metadata capabilities, supporting various transport protocols and authentication methods. It can be distributed as a single binary, Docker images, or pre-built releases, making it adaptable for different deployment needs.

What does MCP Datahub connect to?

MCP Datahub connects directly to DataHub metadata catalogs, specifically supporting versions 1.3 or later. It functions as an MCP server, which stands for Model Context Protocol server. This design aims to furnish AI assistants with structured access to DataHub's extensive metadata capabilities. Its purpose is to bridge AI agents with rich enterprise data contexts, allowing for intelligent interaction with diverse data assets and their associated metadata.

Core features and tools

MCP Datahub exposes several callable tools for AI interaction. These include datahub_search (or search), datahub_get_entity (or get_entities), datahub_get_schema, datahub_browse, get_lineage, get_dataset_queries, document_search, document_search_content, and document_save. These tools are built upon JSON-RPC 2.0 messages for their core protocol. This collection allows AI agents to perform various operations, from searching for entities to retrieving schema information and managing documents.

Who is it for?

MCP Datahub is primarily for AI agents and assistants, as well as developers creating AI-powered tools that require secure and structured access to enterprise data catalogs. It allows AI systems to interact with data context programmatically. Compatible AI clients include Claude, Cursor, Windsurf, Claude Desktop, OpenAI, and chatgpt.com (with mcp-remote for SSE). This is not the right fit if your AI client only supports deprecated SSE transport without mcp-remote.

How to get started

The server authenticates using an API key passed as an environment variable, or through a standard OAuth authorization flow for managed instances. The standalone server uses stdio transport and requires DATAHUB_URL and DATAHUB_TOKEN environment variables.

Use cases

• Conversational Data Discovery: AI agents can find relevant datasets, dashboards, and other entities using natural language search with filters.
• Lineage & Impact Analysis: AI agents can trace data flow at table and column levels to understand data origins and plan for changes.
• Query Analysis & Authoring: AI agents can surface real SQL queries and generate new ones based on actual usage patterns.

MCP Datahub provides a vital connection between AI agents and enterprise data catalogs, enabling sophisticated metadata interactions. Its support for multiple AI clients and flexible deployment options makes it a useful component for AI-driven data initiatives. The tool's ability to expose a range of callable functions enhances AI assistants' capabilities in data discovery and analysis.