What Is DataHub?

What Is DataHub, Really?

  • Before we open the product, let's agree on what DataHub actually is
  • Most people arrive with a mental model borrowed from another tool — we're replacing it before any mechanics are introduced
Wiki BI Tool Spreadsheet DataHub

I'm sure you've run into this.

  • You search Slack for a table name, pick the one that sounds most recent, and build the report. Three weeks later: wrong numbers to leadership. The table was deprecated six months ago and nobody told you.
  • The same problem is happening with AI: an analytics agent queries "revenue_summary" and confidently reports $4.2B in Q3. The metric definition changed -- revenue is now recognized, not booked. The agent doesn't know. Nobody told it.
  • In both cases, the problem isn't missing data. It's missing context.
Slack Search table_v1? table_v2? Which one is current? AI Agent "revenue_summary" → $4.2B Wrong metric Missing context, not missing data

What DataHub is NOT.

  • Not a query engine -- you cannot run SELECT statements in DataHub
  • Not a BI tool -- you cannot build dashboards in DataHub
  • Not a documentation wiki -- it is not Confluence
  • Not just a data catalog -- it's the operational backbone your entire data stack shares: one place where pipelines, governance teams, and AI agents all read from the same context
Query Engine BI Dashboard Wiki Data Catalog Operational Backbone

DataHub is a Context Platform.

  • "A data catalog is a tool. A context platform is the metadata layer your entire stack runs on."
  • Three types of context DataHub unifies:
  • Technical: schemas, lineage, ownership, platform metadata
  • Operational: query patterns, freshness SLAs, quality scores, incident history
  • Business: glossary definitions, governance policies, documentation, runbooks
Technical Operational Business DataHub Context Layer

One Catalog, Two Audiences

  • For you, as a data professional: search and discover assets across all your systems in one place
  • Check lineage before using a dataset in a report; verify data quality before presenting to leadership
  • Know exactly who owns a table and what the fields mean
  • For AI agents in your organization: same context, delivered programmatically via MCP server and API
  • "Ask DataHub" (DataHub Cloud) lets anyone query the catalog in plain English, without knowing table names
  • Governance work has a multiplier: one well-described dataset serves dozens of AI-powered workflows
Analyst Search, lineage, quality, ownership DataHub Catalog AI Agent MCP server, API, Ask DataHub (Cloud)

DataHub builds context from what already exists -- plus what your team adds.

  • Automated ingestion -- connectors crawl Snowflake, dbt, Looker, Confluence, and 80+ other sources on a schedule, plus a plugin that captures lineage directly from Airflow as pipelines run; schema, lineage, ownership, and query history come in automatically
  • Usage & lineage signals -- ingestion also reads query history (e.g. Snowflake, BigQuery) to surface popularity and lineage patterns automatically; DataHub Cloud can auto-draft table and column documentation from that same context
  • Human enrichment -- your team adds the business layer: descriptions in plain English, classification tags, glossary term links, and ownership claims that no automated system can infer
  • Result: a context layer that is always current at the technical level, enriched with organizational knowledge, and available to anyone -- or any agent -- that needs it
Automated Ingestion Usage & Lineage Signals Human Enrichment DataHub Context Layer

Quick Exam