ContextQMD
Back to Datahub

DataHub Agent Context

datahub-agent-context/README.md

1.7.02.3 KB
Original Source
<!-- PyPI long description. Keep concise, feature-discovery-first. -->

DataHub Agent Context

MCP tools for AI agents to search and query your DataHub metadata catalog — works with Claude, Cursor, Copilot, and any MCP-compatible AI assistant.

What you can do

  • Search datasets, dashboards, pipelines, and other data assets by name or description
  • Retrieve entity details — schema, lineage, ownership, tags, glossary terms, and more
  • Trace lineage upstream and downstream across your data assets
  • Mutate metadata — update descriptions, tags, owners, domains, and glossary terms
  • Build LangChain or Google ADK agents with pre-built tool bindings
  • Set up Snowflake AI agents with one CLI command

Installation

bash
pip install datahub-agent-context

# With LangChain support
pip install "datahub-agent-context[langchain]"

Quickstart

LangChain agent

python
from datahub.sdk.main_client import DataHubClient
from datahub_agent_context.langchain_tools import build_langchain_tools

client = DataHubClient.from_env()

# Read-only tools (search, lineage, entity details)
tools = build_langchain_tools(client, include_mutations=False)

# Include write tools (tags, descriptions, owners, etc.)
tools = build_langchain_tools(client, include_mutations=True)

# DataHub Cloud: add Ask DataHub AI assistant
from datahub_agent_context.langchain_tools import build_langchain_cloud_tools
tools += build_langchain_cloud_tools(client, ask_datahub=True)

Snowflake AI agent setup

bash
datahub agent create snowflake \
  --datahub-url https://your-datahub-instance \
  --datahub-token your-token

Available tools

Search — search(), search_documents(), grep_documents()

Entities — get_entities(), list_schema_fields()

Lineage — get_lineage(), get_lineage_paths_between()

Queries — get_dataset_queries()

Mutations — add_tags(), remove_tags(), update_description(), set_domains(), add_owners(), add_glossary_terms(), add_structured_properties(), save_document()

Cloud-only — ask_datahub_chat() (DataHub Cloud AI assistant)