Skip to content
Manuel Luján edited this page Nov 6, 2026 · 2 revisions

LineageIQ wiki

LineageIQ is a read-only lineage, catalogue and governance mapper for Microsoft Fabric. It discovers the workspaces and items it can read, extracts dependencies from notebook code, pipeline definitions and semantic model TMDL down to source objects (tables, views, stored procedures and files), and serves the result as a 2D/3D graph UI, a REST API and MCP servers for AI agents.

It is an early-stage, MIT-licensed project (version 0.2.0). It complements Microsoft Purview and the native Fabric lineage view; it does not replace them.

Try it in one minute

You need Docker and nothing else: no Fabric access and no credentials.

docker run --rm -p 8765:8765 ghcr.io/largolujan/lineageiq:latest

Open http://localhost:8765. The image serves a synthetic, fictional Contoso tenant (three workspaces, 18 items) in read-only demo mode. Try the lineage, reverse impact, DAG and Graph Map views, for example on the RPT_Sales_Dashboard report or the sales_summary table.

To ask an AI agent about the same data, register the stdio MCP server inside the image with your MCP client:

{
  "mcpServers": {
    "lineageiq-demo": {
      "command": "docker",
      "args": ["run", "-i", "--rm", "ghcr.io/largolujan/lineageiq:latest", "python", "backend/scripts/mcp_server.py"]
    }
  }
}

Then ask "what feeds the RPT_Sales_Dashboard report?". More options are in Quick start.

Who it is for

  • Data engineers and platform owners on Fabric who need answers such as "which notebooks write this table?" or "what breaks if this view changes?" before they change something.
  • Governance and data teams who want one cross-workspace view of items, source objects, run history, schedules, workspace roles, sensitivity labels and endorsement.
  • Teams using AI agents that should be able to query lineage and impact through MCP instead of screenshots.

What it does not do

  • No column-to-column lineage: lineage stops at object level.
  • Code extraction is heuristic, so some edges are marked likely or unknown rather than confirmed.
  • The Fabric-native discovery pipeline is assembled by hand.

See How it compares for the full list of limits and the comparison with native Fabric lineage and Purview.

Pages

Elsewhere

Clone this wiki locally