Repository navigation
How It Compares
Microsoft Fabric and Microsoft Purview both show lineage. This page lists the questions where their documented coverage stops and what LineageIQ does instead, so you can decide whether it fills a gap you have. The Microsoft columns reflect Microsoft Learn as of November 2026. Microsoft updates these products often; check the linked sources before you rely on a row.
LineageIQ is meant to sit next to Purview, not to replace it. Purview is the official governance catalogue; LineageIQ covers the object-level and code-level questions Purview does not answer for Fabric items today.
| Question | Native Fabric lineage | Microsoft Purview (Fabric items other than Power BI) | LineageIQ |
|---|---|---|---|
| Which tables does this notebook read or write? | No. The lineage view and impact analysis work with items, not the tables inside them. Materialized lake views have their own table-level lineage view, which covers only those views. | No. Only item-level metadata and lineage are scanned. | Yes, by parsing the notebook code. |
| What breaks if this view or table changes? | Impact analysis lists the downstream items of an item. | No. Lineage stops at item level. | Yes, through the reverse-impact view by source object. |
| Lineage across workspaces? | At item level: the lineage view shows upstream items one level outside the workspace, impact analysis shows downstream items in other workspaces, and the beta Get Upstream Relations API returns related items with their workspaces. | Not supported. | Yes, at item and object level. |
| Notebook to pipeline? | Partly: the beta relations API has an Orchestration relation type, at item level. LineageIQ has not tested this API against a tenant. |
Not supported. | Yes. |
| Lineage of Lakehouse tables and files? | Only through materialized lake views. | Sub-item metadata scanning is in preview; sub-item lineage is not supported. | Yes, as source objects. |
| Can an AI agent query lineage over MCP? | The Fabric Core MCP Server (generally available since September 2026) searches the OneLake catalogue and manages workspaces, items and roles. Its tool list has no lineage tools. | No MCP server for Purview lineage is documented on Microsoft Learn. | Yes: 28 tools across three MCP servers (25 on the local server, 27 on the remote and HTTP servers). See MCP tools reference. |
| Column-to-column lineage? | No. | No: item level only for these items. | No. See the limits below. |
Microsoft Learn pages, checked in November 2026:
- Purview column: How to get lineage from Microsoft Fabric items into Microsoft Purview. Its "Supported capabilities" and "Known limitations" sections state the item-level scope, the preview status of sub-item metadata for Lakehouse tables and files, and that cross-workspace lineage and notebook-to-pipeline lineage are not supported for non-Power BI items.
- Native Fabric column:
- MCP row: Fabric Core MCP Server tools reference and What's new in Microsoft Fabric (general availability, September 2026).
The LineageIQ column comes from this repository: the README, docs/ and the MCP server code in backend/scripts/. You can check every LineageIQ row yourself with the demo image.
- No column-level lineage. Lineage stops at object level: tables, views, stored procedures and files. Columns are catalogued and profiled, but there are no column-to-column edges.
-
Code extraction is heuristic. Lineage is parsed from notebook code, pipeline definitions and TMDL with regular expressions and structured JSON traversal. It can produce false positives and miss dynamic references. Every edge carries a confidence level (
confirmed,likelyorunknown), and review queues list the uncertain cases. - The Fabric-native pipeline is assembled by hand. The notebooks are pasted cell by cell and the Data Pipeline is wired manually. The Fabric-native discovery setup guide walks through it.
-
Identity is by name. A node is
Workspace / Item Type / Item Name; the Fabric GUID is metadata only. Renaming an item is recorded as a removal plus an addition. - Name-based sensitivity matches are hints. They produce false positives and miss opaque column names. Content-based checks cover only the Lakehouse tables the profiling notebooks can reach.
-
Scanner API limits apply.
PostWorkspaceInfoallows at most 500 requests per hour, so governance scans on a large tenant should be scheduled, not run back to back. - No item-level incremental sync. Microsoft's APIs offer workspace-level change polling only, so a refresh rescans changed workspaces.
If your questions stay at item level inside one workspace, or you work mainly with Power BI items, the Fabric lineage view, impact analysis and Purview may already cover what you need. LineageIQ is most useful when you need object-level answers across workspaces, or want an AI agent to answer them.