pbi-doc-generator v1.0)
Turn Power BI Service Workspaces and local .pbix / .pbip / .xlsx files into clean, searchable HTML documentation with interactive Star-Schema diagrams, auto-formatted DAX catalogs, visual Power Query pipelines, 3-panel end-to-end Live Lineage, and a built-in local Model Context Protocol (MCP) server for AI assistants.
👉 Explore the Live Interactive v1.0 Demo Portal in your browser (no installation required)
Click here to download PBI_Doc_Generator_Windows_Portable.zip directly (or view the Release page).
Right-click the downloaded ZIP file and choose Extract All...
Open the extracted folder and double-click PBI_Doc_Generator.exe -> your browser opens http://127.0.0.1:8765 automatically.
(If Windows SmartScreen asks for confirmation on first launch: click More info -> Run anyway.)
pip (Python 3.10+ on Windows, macOS & Linux)pip install "git+https://github.com/djrien-ai/pbi-doc-generator.git"Launch the local Web Studio from anywhere:
pbi-doc-generator(Or clone/download the source repository on Windows and double-click Start_Studio.cmd - it automatically installs requirements.txt on first launch.)
Power BI Service Workspace Role Prerequisite (Contributor or higher)
To extract full semantic model definitions (tables, relationships, DAX measures, and Power Query M) from Power BI Service workspaces, your Microsoft account must have at least the Contributor role (Contributor, Member, or Admin) in each workspace you want to document.
Workspaces where you only have the Viewer role will list basic workspace and report inventory, because Microsoft's Fabric getDefinition (TMDL) API blocks Viewer-only accounts (HTTP 403).
Documenting local .pbix, .pbip, and .xlsx files works 100% offline and requires no Power BI Service permissions.
Related MCP server: fabric-model-reader-mcp
Power BI Service Multi-Workspace Scanner (Zero Azure Admin Setup Required)
Sign in with your normal Microsoft account (Sign in with Microsoft OAuth 2.0 PKCE in your default browser, plus a fallback Power BI Web Window mode for locked-down corporate tenants, Azure CLI, and Service Principal support for CI/CD).
Document 1 or 50+ workspaces into one searchable, 100% offline HTML portal with Grid, List, and By Workspace views, Workspaces & Reports Mapping (highlighting Cross-Workspace Shared Datasets / Thin Reports and Orphan Models), Interactive Live Lineage, and Governance & Health Checks.
Unified Local File Support (.pbix, .pbip, .xlsx)
Drop or select any local .pbix file, .pbip project (automatically resolves its companion .SemanticModel folder), or Excel PowerPivot .xlsx workbook directly in the browser.
Effortless-to-Read Model Documentation
Interactive Star-Schema Diagram: Automatically classifies Fact vs. Dimension tables with zoom, pan, fullscreen, and click-to-pin relationship highlighting (contributed by @jakubpokusa).
Visual Power Query Pipelines: Displays an easy-to-scan transformation step ribbon (1. Source -> 2. Filtered Rows -> 3. Typed Columns) above collapsible syntax-highlighted M code, recognizing 28+ enterprise connectors (SAP HANA, SAP BW, Snowflake, Databricks, Microsoft Fabric, SQL Server, SharePoint, Excel, BigQuery, and more).
Auto-Formatted & Structured DAX Catalog: Defaults to a fast Compact Table overview with a 1-click toggle to Formatted Cards organized by Table -> Display Folder -> Dependency Tier, pretty-printing single-line DAX formulas and linking Reads: / Used by: dependencies.
Interactive 3-Panel Live Lineage Explorer
Click any upstream database/view, Power Query table, column, DAX measure, or Power BI report to trace live upstream origins and downstream impact with smooth SVG connectors.
Built-In Local Model Context Protocol (MCP) Server for AI Assistants
Tab 3 of the Web Studio (pbi-doc-generator mcp) exposes 8 read-only MCP tools to VS Code GitHub Copilot, Claude Desktop, Cursor, or Antigravity across all your scanned workspaces and imported local .pbix / .pbip / .xlsx models, with automatic secret masking.
# Launch the interactive Web Studio (default)
pbi-doc-generator
# Scan specific workspaces incrementally and generate the offline HTML portal
pbi-doc-generator scan --workspace "Supply Chain"
# Generate an anonymized portal safe for external sharing (hides personal/owner names)
pbi-doc-generator portal --anonymous
# Import a local .pbip, .SemanticModel, .pbix, or .xlsx into the multi-model catalog
pbi-doc-generator import-local ./SalesModel.pbip
# Generate a single standalone HTML file for one .pbix / .pbip / .xlsx
pbi-doc-generator file ./Report.pbix
# Start the stdio MCP server for AI assistants
pbi-doc-generator mcp100% Local Execution: Metadata, TMDL definitions, and generated HTML files stay on your machine (127.0.0.1). Zero telemetry or external CDN calls.
Enforced Read-Only API Guard: PowerBiReadOnlyClient blocks DELETE, PATCH, PUT, and mutating POST requests in code, allowing only GET metadata calls and the read-only Fabric getDefinition endpoint.
In-Memory Tokens Only: OAuth access tokens live strictly in Python RAM for the active session and are never written to disk.
Automatic Secret Redaction: Connection passwords, SAS tokens, API keys, and bearer tokens are automatically masked in both HTML pages and MCP responses. See SECURITY.md for details.
The core of pbi-doc-generator is and remains 100% free and open-source under the MIT License. Organizations looking for custom enterprise integrations (automated CI/CD lineage publishing to SharePoint, Confluence, or Microsoft Purview, historical semantic-model drift monitoring, or white-label governance portals) or an independent Power BI Estate & Semantic Model Architecture Audit can inquire via Bohan VOF (connect with Rien Scheerlinck on LinkedIn).
Special thanks to community contributors @jakubpokusa (interactive Mermaid relationship viewer), @okiyuki99 (Xpress9 pbixray support), and @nblankes (TMDL table doc-comment bug report).
If pbi-doc-generator saves your BI team hours of manual documentation or impact analysis, consider starring the repository or supporting ongoing open-source development on Ko-fi.
MIT License - Copyright (c) 2024-2026 Rien Scheerlinck. See LICENSE for full text.
Return high-level counts of scanned Power BI workspaces, semantic models, reports, dataflows, and source connectors.
Inputs: no named parameters.
Regenerate the offline HTML Documentation & Live Lineage Portal from the local catalog store.
Inputs: anonymous.
Return governance, refresh, performance, and architecture findings across all scanned Power BI models.
Inputs: severity.
Return full architecture blueprint of a semantic model: tables, storage modes, relationships, RLS roles, DAX measures, and health findings.
Inputs: model (required).
List documented Power BI semantic models with workspace, refresh status, source connectors, table/measure counts, and health warnings.
Inputs: search, workspace.
Return full upstream (referenced columns and base measures) and downstream (dependent KPIs and reports) lineage for a DAX measure.
Inputs: measure (required), model.
Search across all DAX measures, calculated columns, and Power Query M scripts in the catalog.
Inputs: query (required).
Perform end-to-end impact analysis for an upstream database, server, view, SharePoint site, or table name: returns every affected Power BI table, semantic model, DAX measure, and downstream report.
Inputs: source (required).
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 8 tools
Each tool has a reasonably distinct job: catalog_summary gives counts, list_models enumerates models, and inspect_model drills into one model's blueprint. The only fuzzy pairs are list_models vs inspect_model (depth vs enumeration) and measure_lineage vs source_impact (both lineage, but scoped to measures vs upstream sources), and the descriptions clarify both well enough.
All names are snake_case with a uniform pbi_ prefix and use domain vocabulary consistently. There is a minor deviation between verb-first names (pbi_list_models, pbi_inspect_model, pbi_generate_portal) and noun-first names (pbi_catalog_summary, pbi_measure_lineage, pbi_health_audit, pbi_source_impact), but the pattern remains readable and predictable.
Eight tools is a well-scoped set for a Power BI documentation/lineage server, with each tool covering a distinct read or generate capability. Nothing feels redundant or padded.
The surface covers discovery, inspection, DAX/M search, lineage, impact analysis, health auditing, and portal generation, which is strong coverage for documentation workflows. Minor gaps exist: no standalone report or workspace listing (reports appear only as downstream entities), no per-model export, and no tool to trigger a catalog scan/refresh.
Governed BI MCP. Ask questions of live company data and list workspace sources via OAuth.
Governed data discovery, exact queries, decisions, simulations, and runtime utilities over MCP.
Read-only MCP access to a documented IT fleet: state, changes, posture. 15 tools.
2,000+ MCP servers read at source level. Know what one does before you connect. Free, no key.
