PBIP GA and PBIR Default: What Changes in Power BI

Navy Solv Systems title card reading 'PBIP and PBIR in Power BI' with a folder tree and code file motif.

In Short: PBIP Is Generally Available and PBIR Is Now the Default

Microsoft's Power BI September 2026 Feature Summary, published on 29 September 2026 during FabCon Europe, lists Power BI Projects (PBIP) as generally available and says the Power BI Enhanced Report Format (PBIR) "is now the default report format for PBIX and PBIP files." Put simply, Power BI reports and semantic models are now stored, by default, in a form that can be diffed, reviewed and deployed like code.

What most teams will notice first is that older reports get converted when they are edited. For most estates that is a gain, but a few things are worth checking before someone edits and saves a business-critical report.

What PBIP and PBIR Actually Are

A Power BI Project (PBIP) is a way of saving work from Power BI Desktop. Instead of one binary PBIX file, Microsoft Learn says report and semantic model definitions are "saved as individual plain text files in a simple, intuitive folder structure." A project has two main folders:

The SemanticModel folder holds the semantic model. In the TMDL format (Tabular Model Definition Language), that is a definition folder with separate files for each table, perspective, role and culture, instead of one large model.bim file in the older TMSL format.
The Report folder holds the report. In PBIR, each page, visual and bookmark is its own file in a definition folder, replacing the single report.json file used by what Microsoft now calls PBIR-Legacy.

The .pbip file at the root is optional and just points to the report folder. A default .gitignore excludes the local data cache (cache.abf) and localSettings.json, so data stays out of the repository.

The report format is the real shift. According to Learn, the legacy report.json "doesn't support external editing", while PBIR "is a publicly documented format that supports modifications from non-Power BI applications", with a public JSON schema for every file. PBIR is the default inside PBIX files too, but the source control benefits only come once you save as PBIP.

What "Generally Available" and "Default" Mean in Practice

The feature summary puts it briefly: "New reports use it from the start. When you edit a report that uses the legacy format, Power BI tools automatically convert it to PBIR." Learn's PBIR page goes further: "When you edit and save a PBIR-Legacy report, Power BI silently and automatically converts it to PBIR." Reports in PBIR-Legacy still open in Power BI Desktop and the service. They only change when someone edits and saves them.

Before converting a report, Power BI creates a backup:

Power BI Desktop keeps the backup for 30 days, in a TempSaves Backups folder under the user profile.
The Power BI service keeps a PBIR-Legacy backup for 28 days. You can restore it from Report settings with Restore as PBIR-Legacy, but only for reports converted directly in the service.

There is an important caveat about going back. Learn says: "Restoring a PBIR-Legacy backup doesn't prevent another conversion. To keep a report in PBIR-Legacy format, don't edit it in the Power BI service, or use a release of Power BI Desktop released before September 2026." In practice, plan for conversion being one-way.

Switching between PBIX and PBIP is separate and works both ways through File, Save as. One adjacent feature is still in preview: detecting and reloading external PBIP changes in an open Power BI Desktop session.

Why It Matters: Source Control, Review and CI/CD

Learn says PBIR "greatly improves change tracking and merge conflict resolution". Because each page, visual and table sits in its own file, a pull request shows exactly what changed, and two developers working on different pages no longer collide in one huge JSON file. It also makes scripted work practical, such as copying pages between reports or batch edits across every visual.

For release management, Learn says Fabric Git integration and the Fabric REST APIs export report definitions using PBIR. That is the foundation for Git integration and CI/CD in Microsoft Fabric and for promoting content through Fabric deployment pipelines. One detail is easy to miss: publishing a project from Power BI Desktop uploads a temporary PBIX including the local data cache, while Git integration and the APIs deploy metadata only. For where authoring should happen, our guide to Power BI Desktop vs Power BI Service covers the split.

What to Check Before Converting

These all come from the limitations Microsoft Learn lists for projects and PBIR:

Sensitivity labels - "Sensitivity labels aren't supported with Power BI projects." This may decide which reports move first.
Power BI Report Server - Learn says Power BI Projects are not currently supported in the version of Power BI Desktop optimised for Report Server. Check the current position on Learn before changing that estate.
Where files live - you cannot save a PBIP directly to OneDrive or SharePoint, and a locally synced folder may cause sync problems.
Path length - Windows limits paths to 260 characters by default. Keep the project root short.
Custom visuals - private visuals loaded from a pbiviz file are stored in the report's CustomVisuals folder, while AppSource and organisational visuals load automatically. Test visual-heavy reports after conversion.
Report behaviour - visual automatic filters are written to the PBIR file only once the filter pane has been expanded during editing. Report page synonyms (the linguistic schema) are not supported with projects.
Size - the service enforces a maximum of 1,000 pages per report, 1,000 visuals per page and 300 MB across all report files.
Tooling discipline - save externally edited files as UTF-8 without BOM, set Git's autocrlf, and do not edit undocumented files such as report.json outside Power BI Desktop.

The Power BI Authoring MCP Server: Remote in Preview, Local Generally Available

PBIP matters more now that AI agents edit Power BI content. The September summary announces a Remote Power BI Authoring MCP Server (Preview) at https://api.fabric.microsoft.com/v1/mcp/powerbi/authoring. Microsoft says it lets AI applications create and modify semantic models in Fabric workspaces "without installing or maintaining anything local", and that it "unblocks Mac users ... because the server runs remotely and does not require local binaries." Learn calls this the hosted option. It needs a Fabric tenant setting turned on, and it does not support transactions or trace tools.

The local server, "released in November 2025, remains generally available." You need it to work against Power BI Desktop, PBIP or TMDL files on disk. It is the server we covered as the Power BI Modeling MCP server. Its npm package and GitHub repository still use the powerbi-modeling-mcp name. Learn advises against registering the local and remote servers at the same time.

Two other tools are worth knowing. Power BI Desktop Bridge lets an agent reload changes in a running copy of Power BI Desktop and take screenshots to check its own work. Learn lists the Bridge as in preview. The powerbi-authoring plugin in microsoft/skills-for-fabric bundles semantic modeling and report authoring skills with the local server. Microsoft says it now supports report authoring for remote semantic models and custom visuals. The authoring server itself performs modelling operations only; report pages are edited by skills working on PBIR files. That is why Learn's advice to "Work in PBIP files under Git" matters: every change an agent makes appears as a diff you can review and revert. For how this fits with the other servers, see the Fabric MCP stack.

A Practical Migration Checklist

1. Inventory first - flag reports with sensitivity labels, Report Server dependencies, private custom visuals or very large page counts. 2. Standardise Desktop - get the team onto a current release so everyone saves in the same format. 3. Pilot one report - edit and save it, confirm it converted to PBIR, and note where the backup sits. 4. Save as PBIP into Git - use a short local path, keep the default .gitignore and set autocrlf. 5. Choose a deployment route - Fabric Git integration, the Fabric APIs or Desktop publish, knowing that only Desktop publish carries the data cache. 6. Set review rules - pull requests for every change, with named approvers for model changes and report changes. 7. Decide on agent tooling - the hosted server for models in Fabric workspaces, the local server for Desktop and PBIP work, never both registered at once. 8. Tell your authors - editing a legacy report converts it, including in the service.

Where Solv Systems Comes In

The format change is the easy part; the working model around it - repository structure, branching, review and release - is what pays off. Our Power BI consulting team sets up PBIP projects, Git workflows and AI-assisted development so they hold up once more than one developer touches a model.

Sources and Further Reading

Frequently asked

PBIP (Power BI Project) is a way of saving Power BI Desktop work as a folder of plain text files instead of a single PBIX file, with one folder for the semantic model and one for the report. PBIR (Power BI Enhanced Report Format) is the format of the report definition itself, with each page, visual and bookmark in its own JSON file. Semantic models in a project use TMDL.

Yes, when you edit them. Microsoft Learn says that when you edit and save a PBIR-Legacy report, Power BI silently and automatically converts it to PBIR. Existing PBIR-Legacy reports still open in Power BI Desktop and the service without changing until they are edited and saved.

Only in a limited way. Power BI Desktop keeps a backup for 30 days, and the service keeps a PBIR-Legacy backup for 28 days that you can restore from Report settings, for reports converted in the service. Learn notes that restoring a backup does not prevent another conversion, so in practice treat the change as one-way.

No. PBIR is the default report format for both PBIX and PBIP files, and Power BI Desktop can save a PBIX as a PBIP and a PBIP as a PBIX through File, Save as. The source control benefits come from saving as PBIP, because that is what turns the report and model into text files Git can track.

Check sensitivity labels, which Learn says are not supported with Power BI projects; Power BI Report Server, whose version of Power BI Desktop does not currently support projects; where files are saved, because PBIP cannot be saved directly to OneDrive or SharePoint; the Windows 260-character path limit; private custom visuals; and very large reports against the PBIR size limits.

The local Power BI Authoring MCP server, which is generally available, can work against PBIP and TMDL files on disk. The remote, Microsoft-hosted version is in preview and works against semantic models in Fabric workspaces with nothing to install, which is why it works on a Mac. Microsoft Learn recommends working in PBIP files under Git so every change an agent makes can be reviewed and reverted.