Advanced NL2SQL and advanced DAX generation in Fabric data agents (Generally Available)
Data agents in Microsoft Fabric let users ask questions about data in plain language. As a creator, you connect an agent to supported sources—including SQL databases, lakehouses, Power BI semantic models, and eventhouses—add the business context it needs, and share it with users across Fabric, Teams, Microsoft 365 Copilot, Copilot Studio, and Microsoft Foundry. The data agent can then respond like a domain expert that understands what your tables mean, how your metrics are defined, and which source can answer a question. The usefulness of a data agent ultimately depends on a simple question: can it translate what a user asks into the right query? That is harder than it sounds. Enterprise schemas are rarely intuitive. Several tables, columns, measures, or join paths may appear to answer the same question, so a query can be technically valid but still return the wrong result. Advanced NL2SQL and advanced DAX generation help address this problem by allowing the agent to explore the schema and data, validate assumptions, and refine its approach before generating a query. Both capabilities are available today, and we'll begin rolling the core improvements out to the standard runtime on November 2, with the rollout expected to complete by November 10. In this post, we'll look at how they work, what our internal benchmarks show, the new configurations available for shaping how the agent behaves, and what to consider as you re-test and evaluate your data agents. Why conversational analytics gets harder with enterprise data Turning a plain-language question into a correct query is a complicated problem, both for SQL and DAX, requiring the agent to resolve several challenges at once: Schemas are large. Enterprise sources can contain hundreds of tables, columns, and measures, and the names alone may not explain when or how they should be used. Business context exists outside the data. Definitions such as “revenue” or “active customer” often live in documentation, previous queries, or analysts’ knowledge rather than in the source schema. Filter values must match the way data is stored. Users describe concepts naturally, but those concepts must map to the values in the data. For example, a user may ask about “California” when the relevant column stores the value “CA.” Questions are ambiguous. More than one column, calculation, or join may appear to satisfy a request. Individually, each of these can be managed with enough hand-written guidance. The difficulty is that they compound: a single-pass generator has to resolve all of them at once — your schema, your definitions, your filter values — committing to an interpretation before it has any evidence that interpretation is right. When it guesses wrong, the query still runs, and the answer looks no different from a correct one. Advanced NL2SQL and advanced DAX generation are built around that problem. Rather than generating a single pass, the agent gathers evidence first. Advanced NL2SQL: the agent explores before it writes The shift is architectural. Instead of one-shot generation, the agent has sub-tools to explore schema, test joins, and look up real filter values before it writes SQL. Under the hood, you can inspect each step the agent took, and the assumptions and reasoning behind it. It operates in three modes, each backed by its own sub-tools: Mode What it does Example query following Finds a relevant example query and answers by making approved modifications, such as literal filter or parameter substitutions. Clarifying questions Determines it can’t answer reliably without more information and asks targeted clarifying questions before generating SQL. Data exploration When no example directly applies, learns from existing example queries to infer business logic, explores the data to validate assumptions, and iteratively refines its approach. What you get as a creator: Advanced NL2SQL explores the SQL source schema, tests joins, and looks up real filter values before generating SQL, helping it choose the appropriate tables, columns, relationships, and filters. Schema descriptions and topics that teach the agent what your data means, so you get better answers with fewer instructions. Better example-following and filter-value substitution for novel questions. Handles large schemas far better and runs faster end to end, with more reasoning detail for debugging. New configurations: schema descriptions and topics Two new configurations give the agent more of the context it needs. Both are available for SQL data sources on the preview runtime. Schema descriptions Schema descriptions attach business context to the objects in your data source — what a table holds, what a column represents, and how each should be used. The agent uses them to interpret large or ambiguous schemas and generate a more accurate query. They matter most where object names don’t carry meaning on their own, or where several tables and columns look interchangeable. Figure: Schema descriptions add business context to schema objects, helping the data agent interpret your data. Topics Topics help you give the agent detailed guidance without sending every instruction with every question. The agent searches your topic content and uses only the sections relevant to the current request. Use data source instructions for rules that must always apply, and use topics for deeper guidance that applies only to a specific subject. You can write topics in Markdown, import multiple files, and review the run steps to see which topics the agent selected. Figure: Topics organize subject-specific instructions in Markdown to guide the data agent’s query generation. Internal benchmark testing: Advanced NL2SQL on Preview Runtime vs NL2SQL on Standard Runtime Better reasoning matters only if it produces better answers without making users wait substantially longer. In an internal benchmark set evaluating answer quality and end-to-end latency, the preview runtime delivered the stronger outcome, answering more questions correctly than the standard runtime while reducing both median and tail latency. Smart-answer accuracy increased from 78.8% to 90.9%. Latency improved at the same time: P50: 28.4 seconds to 26.0 seconds P95: 46.8 seconds to 32.9 seconds Figure: Smart answer accuracy and end-to-end latency, standard runtime versus the Advanced NL2SQL preview, on an internal benchmark set. Accuracy rises 12.1 points to 90.9%, while P50 latency falls 8.3% and P95 falls 29.8%. *Performance results are based on Microsoft internal testing conducted in 2026 using an internal evaluation set comparing standard NL2SQL to the advanced NL2SQL preview in Microsoft Fabric. Actual results may vary depending on workload, data model, and configuration. The P95 result shows the largest latency improvement in this benchmark, with latency decreasing from 46.8 seconds to 32.9 seconds. This improvement is most relevant to the slower, more complex queries represented at the tail of the distribution. Advanced DAX generation: agentic reasoning over semantic models The agent calls tools, inspects results, and refines its approach to answer harder questions. Instance value indexing searches real column values, so filters match your data. And you get consistent results across Data Agent, Power BI, and M365 Copilot. For users, this means the agent is more likely to resolve filter terms to the values stored in the semantic model and return consistent answers across supported Microsoft experiences. Internal benchmark testing: Advanced DAX generation improves accuracy while cutting latency Our internal Advanced NL2DAX evaluation showed improvements in both answer accuracy and latency. Smart-answer accuracy increased from 89.15% to 91.86%. At the same time: P50 latency: 39.1 seconds to 20.4 seconds P95 latency: 97.5 seconds to 34.7 seconds This represents an approximately 48% reduction in P50 latency and a 64% reduction in P95 latency, along with a 2.7 percentage-point increase in smart-answer accuracy. *Performance results are based on Microsoft internal testing conducted in 2026 using an internal DAX generation evaluation set comparing the baseline to the advanced DAX generation preview in Microsoft Fabric. Actual results may vary depending on workload, semantic model, and configuration. Rolling out the runtime improvements We'll begin rolling the core Advanced NL2SQL and advanced DAX generation improvements out to the standard runtime on November 14, with the rollout expected to complete by November 10. As part of this rollout, the standard runtime will also move to GPT-5.6 Terra. As the rollout reaches your environment, data agents using the standard runtime will receive these improvements and the model upgrade by default, with no opt-in required. What is moving to the standard runtime: Advanced NL2SQL: improved example-query following, filter-value substitution, and data exploration. Advanced DAX generation: the agentic tool-calling loop and instance value indexing. Model upgrade: The standard runtime will move to GPT-5.6 Terra for orchestration and supported data agent capabilities. What stays in preview: Schema descriptions and topics for SQL sources. We believe these changes deliver significantly better results, based on improvements observed in our internal evaluations. However, results can vary by data source, model, instructions, examples, and workload. Creators can re-test and evaluate their agents on the standard runtime between November 14 and November 20 as the rollout reaches their environment, or try the changes now on the preview runtime. Moving from one-shot generation to an agent that explores data can change how existing instructions and examples are interpreted. We recommend re-testing: Run your existing question set against the standard runtime as the rollout reaches your environment between November 14 and November 20, or switch to the preview runtime to try the changes now. Rerun your evaluations with the Fabric data agent SDK and a ground-truth set. Compare the new results with your previous baseline, and use the visible reasoning steps to identify where the agent’s assumptions differ. Tune your example queries and instructions based on what the evaluation shows. The agent now explores your data and adapts examples through substitution, so guidance written to work around the previous behavior may no longer be needed. Re-testing and evaluating your agents will help you confirm the expected quality improvements for your own data and scenarios. Next steps If you build data agents in Microsoft Fabric, re-test your agents on the standard runtime as the rollout reaches your environment between November 14 and November 20, or try the changes now on the preview runtime. For more details: Fabric data agent runtime: Learn what the standard and preview runtimes each contain, how to switch between them, and what’s in the preview runtime today. SQL sources in Fabric data agent: Learn how NL2SQL and Advanced NL2SQL generate queries over Lakehouse, Data Warehouse, SQL Database and Mirrored Database sources, and which configurations each one can use. Semantic model best practices for data agent: Semantic model best practices for data agent: Learn about advanced DAX generation and how to prepare a semantic model so the agent can answer questions accurately.AI @Work: Understand AI, Work with Agents, Build Your Skills
Curious about AI and agents, but unsure where to start—or what they mean for your job? Join us for an introductory session **delivered in Arabic**, designed for both **IT and non-IT audiences**. We’ll explore AI in everyday work, explain agents in simple terms, and demonstrate how to communicate with them effectively. Through practical examples and a simple agent-building demo, we’ll connect these concepts to familiar tasks and show why your knowledge, judgment, and skills matter when working with AI. What we’ll cover: Why understanding AI matters, whatever your role. What AI agents are and how they can help with everyday tasks. How to write clear prompts, provide context, and evaluate responses. How to turn a simple idea into an agent using instructions in Copilot. How agents fit into workplace tools, including an example from Microsoft Fabric. How AI may change tasks and roles, and which skills can help you adapt. Who should attend? Business professionals, data practitioners, IT professionals, students, and anyone interested in understanding and using AI at work. By the end of the session, you’ll have a clearer understanding of AI agents, practical ways to interact with them, and a starting point for building your skills. Level: Introductory — no coding or previous AI experience required.From a prompt to an operational dashboard: Generate Real-Time Dashboards with Copilot (Preview)
Start with the question, not a blank dashboard Which transit stations are sending the most fans toward a stadium? How does that incoming demand compare with activity at the gates? Where are queues building? Answering these operational questions traditionally requires translating them into tables, queries, visualizations, and a dashboard layout. When you already know what you want to monitor, that setup can get in the way of the question you are trying to answer. Copilot in Microsoft Fabric can generate a Real-Time Dashboard from a natural-language description of the insights you want. Working with a selected database in an Eventhouse, it uses the tables in that database to create a dashboard tailored to your goal. Figure: Start generating a Real-Time Dashboard with Copilot from an Eventhouse database. Rather than beginning with an empty canvas and building each visual separately, you describe the operational view you need. Copilot helps create the initial dashboard, and you review and customize it for your team. Describe, refine, and create Copilot can already help you write KQL queries and edit individual dashboard tiles. With dashboard generation, you can start one level higher: describe the operational view you need and let Copilot bring several related insights together into an initial dashboard. Copilot works from the context you provide about the dashboard you want to create. Describe the operational goal first, then add the details that matter for the view you have in mind. For example, specify the signals you want to monitor, the comparisons that are important, and any tables you know are relevant. You can also describe how you want the information organized. You might ask for an overview of the most important measures together with trends over time and breakdowns by location. You don't need to define every tile before you start. Use follow-up messages to clarify the request as Copilot develops the dashboard plan. Before creating the dashboard, review how Copilot has interpreted the request. This gives you an opportunity to narrow the scope, emphasize a particular comparison, or add context that was missing from the original prompt. From your Eventhouse, open New Real-Time Dashboard and select Generate with Copilot (Preview). In the Copilot pane, describe your goal, the data you want to prioritize, and how you want the dashboard organized. Data-specific example prompts can help you get started. Figure: Describe the operational insights you want in the Copilot pane. The experience is conversational. You can follow how Copilot interprets your request and use follow-up messages to refine the scope or focus before creating the dashboard. For example, you might clarify which tables matter or ask to emphasize comparisons between locations instead of an overall summary. When you're ready, select Start dashboard creation, choose a dashboard name and workspace location, and select Create. The result is a Real-Time Dashboard you can continue to edit, not a static image of suggested charts. Figure: Choose a name and workspace for the generated Real-Time Dashboard Example: Connect transit arrivals with stadium gate load Consider a stadium operations team preparing for an event. Transit arrivals provide one view of incoming demand, while gate activity and queue lengths provide another. Looking at those signals together helps the team understand where to focus its attention. In this example, the prompt points Copilot to the relevant transit, stadium gate, stadium zone, and match-event data. It asks for estimated passengers and fans heading to the stadium, broken down by station and transit line, alongside gate entries and queue lengths. The goal is specific: compare incoming transit demand with activity at stadium entrances. That level of context makes the desired analysis concrete. The goal isn't simply to visualize every available column. It's to compare people arriving by transit with activity at the stadium entrances. The generated dashboard brings several views together: Summary tiles for estimated transit passengers, fans heading to the stadium, the relationship between those totals, and average gate queue length. A comparison of transit demand and gate entries over time. Passenger breakdowns by station and transit line. Gate-level and zone-level comparisons of transit inflow and gate activity. A detailed table of current gate load and queues. Together, these views give the team an overview and a way to inspect individual locations. A summary can suggest where to look next; a station breakdown or gate table provides more detail for that investigation. As with any generated dashboard, review the result before relying on it for operational decisions. Verify the queries, relationships between the data, and the meaning of the measures to make sure the dashboard represents the intended analysis. Make the dashboard your own An initial dashboard gives you something concrete to evaluate. Are the most important measures easy to find? Does each chart answer a clear question? Would a different visual make a comparison easier to understand? After generation, you can change visual types and modify the queries behind the tiles. Copilot can also help with editing visuals or writing queries as you refine the dashboard. Figure: The generated dashboard combines transit demand, gate activity, and queue information. Next steps Explore Generate Real-Time Dashboard using Copilot for the workflow and use Copilot's feedback controls to share what works and what needs improvement. Learn how to further Customize Real-Time Dashboard Visuals to make this dashboard your own.Data Days: Learn, build, and connect with AI @Work
Real people. Real projects. Real AI on Fabric, SQL, and Power BI. From November 2 through December 15, Data Days is your chance to get past polished demos and put data and AI to work yourself. Join a live session, build something new, work toward a certification, or connect with a community study group. Share what you've learned, showcase something you've created, or bring Data Days to your own community. This is about doing, not just watching. Start with the series that interests you AI Work is at the heart of this Data Days, with three stages designed to meet you wherever you are: Let's Make Sense of It explores how Copilot, Fabric IQ, Fabric Apps, and your data fit together. If you've heard the names but want to understand what they mean for your work, start here. Let's Build Something gets hands-on. Watch practitioners build Fabric Apps, work with Spark to make data AI-ready, and turn ideas into something useful. See It in Action showcases real projects, including Fabric IQ and Fabric Apps examples. Discover what community builders created, the decisions they made, and what they learned along the way. There's more to explore, too. Catch up on announcements and demos with the FabCon & SQLCon Recap series, prepare for certification with Let's Get Into Certifications, or dive into Power BI data visualization sessions. Learn, create, and connect with the community Data Days goes beyond live sessions. Join a community study group, explore the Fabric Apps Gallery, or put your creativity to the test in the Power BI Dataviz World Championships and community contests. Have something to share? Host a session, lead a study group, or bring an existing community event into Data Days. Whether you're learning, building, or teaching, there's a place for you. Want to work toward a certification? Data Days includes exam preparation, community study groups, and limited certification voucher offers. Eligible participants can explore 100% discount offers for DP-600, DP-700, and DP-800, plus 50% discount offers for select additional exams. Eligibility requirements and deadlines apply. Come with curiosity. Leave with something you want to try, someone new to learn with, or something you've created yourself. Explore Data Days today!Power BI September 2026 Feature Summary
As the Power BI, Microsoft Fabric, data, and analytics community gathers in Barcelona for the European Microsoft Fabric + SQL Community Conference, this month is packed with news. From new Copilot and AI experiences to reporting, modeling, data connectivity, developer enhancements, and mobile updates, we're sharing the latest innovations designed to help you build, analyze, and act on your data more effectively. Whether you're joining us in person or following along from afar, this roundup is your guide to everything new across Power BI this month. General Deprecation of Old File Picker experience in Power BI Desktop In April we deprecated the old file picker experience in Power BI Desktop. Starting in October, users using versions of Power BI Desktop from March 2026 or earlier will no longer be able to save and share Power BI files to OneDrive and SharePoint. Please ensure you update Power BI Desktop to keep using this functionality. Copilot and AI Power BI Agentic Experiences (Generally Available) Agentic development in Power BI is real, and it is already delivering value: whether you are starting from scratch with greenfield projects or building upon existing solutions in brownfield scenarios, BI developers can delegate the work to an agent, review the result, and refine it rather than starting from scratch every time. What makes this work is tools and context. Through Work IQ and Fabric IQ, the agent grounds its work in your business context—your data, your semantics, and the way your organization operates. Through skills—published by Microsoft in skills-for-fabric, shared by the community, or personalized by your own team—it also picks up your development context: the modeling and reporting practices, conventions, and workflows your team expects. Through tools such as the Power BI Authoring MCP server and the Power BI Desktop Bridge CLI, the agent can act on that context: reading and changing semantic models and reports directly, reloading its edits in Power BI tools, and reviewing its own work. Point an agent at a requirements document or a screenshot of the layout you want, describe what you need, and the agent can help accelerate the initial authoring pass, which you can then review and refine. rt authoring skills. What's available today The powerbi-authoring plugin in microsoft/skills-for-fabric bundling the semantic modeling and reporting agent skills and tools. It was recently updated to support report authoring for remote semantic models in Power BI service and an expanded range of supported report visuals, including custom visuals. To learn more, refer to the Power BI skills documentation. The local Power BI Authoring MCP server allows the agent to edit semantic models in Power BI Desktop. To learn more, refer to the Power BI Authoring MCP documentation. Power BI Desktop Bridge lets an agent reload changes in a running instance of Power BI Desktop and capture screenshots to check its own work. To learn more, refer to the Power BI Desktop Bridge documentation. Browser-based validation for remote report editing, so the agent can validate a report using a headless browser (an automated browser that runs without a visible user interface). Removing the local Power BI Desktop dependency allows the full authoring loop to run without Power BI Desktop, extending the agentic experience to Mac users and others who do not rely on a local installation. To learn more and get started, refer to the Power BI agentic documentation. Microsoft 365 Copilot is coming to Power BI (Preview) Copilot experiences have emerged across the applications people use every day, bringing AI closer to the work happening in each product. But this can also feel fragmented, with different entry points, separate conversations, and uncertainty about which Copilot agent to use. Our goal is to create a consistent Copilot experience that is available wherever people work. People should not have to leave the application they are using to find it. Microsoft 365 Copilot should be available within their experience, carry relevant context across experiences, and connect to the specialized intelligence of each product. We’re taking an important step toward that goal with a preview of Microsoft 365 Copilot integrated directly into Power BI and Microsoft Fabric. This integration gives users a familiar place to bring together business context and trusted data. Shared conversation history lets them continue a discussion as they move between Microsoft 365 Copilot experiences instead of starting over in each application. ower BI experience, including integration with Work IQ. In Power BI, a question no longer has to stop with the numbers in a report. A sales leader could use Fabric IQ to identify which regions are behind plan, then use Work IQ to connect the results with customer emails, meetings, account plans, and documents. They could summarize the complete picture and even kick off actions and tasks in Cowork, all from inside Power BI. This work context and ability to move directly from insight to action were not available in the standalone Power BI Copilot experience. the Cowork integration into the Power BI experience, highlighting the full Microsoft 365 experience embedded into Power BI. A new setting in the Fabric admin portal will determine whether this integration is available by enabling or disabling its entry point in the navigation pane. It will be off by default, giving administrators time to evaluate the experience before introducing it to their organization. Administrators can enable it when they are ready and disable it if they need to pause access. This setting is independent of the existing setting for standalone Power BI Copilot. If both settings are enabled, users with a Microsoft 365 Copilot license will see entry points for both experiences, and standalone Power BI Copilot will continue to appear on Home. If standalone Power BI Copilot is disabled and Microsoft 365 Copilot is enabled, users will see only the Microsoft 365 Copilot entry point. During the initial preview, the integrated experience is available only to users with a Microsoft 365 Copilot license. These users receive the full Copilot Chat experience in Power BI and Fabric, bringing together Fabric data through Fabric IQ and relevant Microsoft 365 context through Work IQ. Existing permissions and governance controls continue to determine which content each user can access. Support for a broader Fabric audience will follow. This expansion will give Fabric users without a Microsoft 365 Copilot license access to an integrated Copilot experience tailored to their available capabilities. The experience may differ from the full licensed offering, but it will extend Fabric IQ and conversational data experiences to more types of Fabric users. The preview will roll out with the setting turned off by default, with administrators choosing when to enable it for their organization. This is an important step toward one Copilot that meets people wherever they work, connecting the context surrounding each decision with trusted business data. To learn more, refer to the Copilot in Power BI documentation. Power BI data answers in Microsoft Copilot Chat (Generally Available) Users can find and ask questions about Power BI reports and semantic models directly from Microsoft 365 Copilot. They can search for relevant Power BI content, attach a report or provide a link, and ask questions in natural language. Powered by Fabric IQ, the experience uses the data, measures, relationships, and business logic already defined in Power BI to provide grounded answers. Users can also bring Power BI insights into conversations that include their Microsoft 365 work context, such as emails, meetings, and documents. Microsoft 365 Copilot now supports Microsoft Purview Data Loss Prevention (DLP) policies and information protection controls for this experience. This helps organizations adopt conversational analytics while applying their existing policies for identifying, classifying, and protecting sensitive data. The experience continues to respect existing Power BI permissions and Microsoft 365 governance controls. Users can only access Power BI content they are authorized to view, and administrators retain control over whether Fabric data is available in Microsoft 365 Copilot. Organizations can use Power BI data answering in Microsoft 365 Copilot with production support and service commitments. This milestone makes it easier for users to bring trusted business data into the flow of work without leaving Microsoft 365 Copilot or creating a separate agent. To learn more, refer to the Fabric IQ in Microsoft 365 Copilot Chat (Frontier) documentation. Fabric IQ Plugin in Microsoft 365 Copilot Cowork (Generally Available) Power BI data is available where multi-step work happens, so a business question, the analysis behind it, and the follow-up it triggers stay in one place. To ground a task in Power BI, users can search for relevant content, attach a report, or paste a link to a report or semantic model. Questions are answered using the data, measures, relationships, and business logic already defined in the semantic model. Coverage now extends to reports published through organizational apps, with improvements to content discovery, large model handling, schema selection, Verified Answers, and row-level value search. The plugin also carries forward the governance controls organizations rely on in Power BI: Existing access controls: Power BI permissions and row-level security determine which reports and semantic models each user can reach. Citations and sensitivity labels: Responses cite the Power BI content behind an answer and surface the Microsoft Purview Information Protection label applied to it, so users can see where a number came from and how it is classified. From insight to action: With Fabric IQ, users can analyze Power BI reports using natural language, uncover key findings, and put those findings to work right away in Cowork. Whether they are summarizing business performance, investigating a trend, or answering a data question, Fabric IQ turns trusted organizational data into outcomes. To learn more about supported scenarios, setup, and current considerations, refer to the Fabric IQ in Microsoft 365 Copilot Cowork documentation. For licensing and pricing information, refer to Microsoft 365 Copilot plans and pricing. Build Fabric Apps more easily in GitHub Copilot App (Preview) Fabric Apps were introduced in June as an AI-first way to build custom web apps with Microsoft Fabric as the backend. This month we’re excited to share that building these Fabric Apps on your governed data has become even easier with the GitHub Copilot App, which gives you a natural language chat interface instead of a command-line interface. . To try it out, start by downloading the GitHub Copilot App. Then navigate to Customize > Plugins > Rayfin, and select Install to enable our dedicated plugin: Rayfin plugin in GitHub Copilot App. Once the plugin is enabled, simply start a new chat and work with Copilot to help plan your Fabric data app. After you review and approve the plan, ask the agent to execute it and specify the workspace where you want to deploy the app. . We look forward to seeing the amazing apps you build! Check out our community gallery for inspiration. Enhanced Report Consumption Copilot with Fabric IQ When viewing a Power BI report, you can enable Fabric IQ from the Copilot pane to better answer complex questions about your data. This enhancement merges existing Fabric IQ capabilities into Power BI Copilot report viewing mode, allowing Copilot to use a multi-step approach to evaluate data in the semantic model and refine its approach when generating a response. . With Fabric IQ enabled, Copilot can: Provide better answers to questions about your data based on detailed reasoning capabilities directly into the report viewing experience. Show reasoning steps to help you understand how Copilot provided an answer. Generate Fabric visuals to help you further analyze your data. Answers can include column, bar, stacked, line, area, pie, donut, scatter visuals. Provide the underlying DAX query and data table used to generate a response. . To learn more, refer to the Ask Copilot questions about your data documentation. New control for Copilot access to semantic models Power BI now provides semantic model editors with greater control over how Copilot can be used with their data. Users with write permissions can block read-only users from accessing a semantic model through Copilot and Microsoft’s remote MCP tools, supporting governance, compliance, and access-control requirements. The setting applies across Copilot experiences that query the semantic model, including Power BI Copilot experiences; Microsoft 365 Copilot Chat and Cowork; and data agents, remote Fabric MCP servers, and Power BI MCP servers. When disabled, users with read-only access won't be able to access that semantic model through these experiences. By default, the setting remains enabled to preserve the current behavior. This setting appears in the Copilot section with the following description: “Allow any person with read-only access to use AI (i.e. Copilot, Microsoft’s MCP tools) on the model and its related reports.” Setting that allows people with read-only permission to use AI experiences with the semantic model and its related reports. This update gives organizations a clear and enforceable way to control AI access to semantic model data while maintaining flexibility for semantic model editors. To learn more, refer to the Use Copilot with Semantic Models in Power BI documentation. Remote Power BI Authoring MCP Server (Preview) The Power BI Authoring MCP enables AI applications to create and modify semantic models in Fabric workspaces, without installing or maintaining anything local. This release also unblocks Mac users that want to use the Power BI Authoring MCP, because the server runs remotely and does not require local binaries. Use the following endpoint in AI applications such as GitHub Copilot App: https://api.fabric.microsoft.com/v1/mcp/powerbi/authoring The local Power BI Authoring MCP server, released in November 2025, remains generally available. It remains an option when an agent needs to work with Power BI Desktop, Power BI Project (PBIP) or TMDL files on disk. The local and remote servers support the same authoring capabilities. We recommend using the remote server when local development is not required, such as when creating a Direct Lake semantic model in a Fabric workspace. In the following example, the AI agent uses one tool call to create a Direct Lake semantic model connected to a Fabric Lakehouse: To learn more, refer to the Power BI Authoring MCP server documentation. Reporting New Report Ribbon on the web (Preview) The Report ribbon is now available on the web, providing a consistent experience for report consumption and authoring that more closely matches Power BI Desktop. The ribbon organizes commands into clear, easy-to-navigate tabs, making it easier for users to discover features, interact with reports, and complete editing and authoring tasks directly in the browser. It also introduces the new Design ribbon and additional capabilities that were previously unavailable on the web, such as Format Painter. This milestone brings greater consistency to Power BI report-authoring experiences and marks an important step toward a seamless, end-to-end reporting workflow on the web. As web authoring capabilities continue to expand, users will be able to accomplish more of their reporting tasks directly in the browser, enabling a more streamlined, productive workflow. Enable the Try the new ribbon (preview) at the top of the page to explore the new experience. You can disable the setting to return to the previous ribbon. To learn more, refer to the Report Ribbon on the web documentation. Live Visual Sharing in Outlook on the web (Preview) For years, sharing data through email has been stagnant. You could send a screenshot, but it immediately became stale. You could attach a report link, but recipients had to leave their workflow to find the context they needed. And every follow-up question often meant another screenshot, another export, or another email thread. Now, we're bringing a new way to share Power BI insights directly in the flow of work with Live Visual Sharing in Outlook. With Live Visual Sharing, you can insert a Power BI visual into Outlook on the web and share it as a live, interactive experience. Recipients can view and interact with the visual directly in Outlook while staying connected to the latest underlying data. When a live visual is shared, Outlook renders an interactive version of the Power BI visual. Rather than embedding a static image, the experience remains connected to the source report and retrieves current data when viewed. Recipients can: Interact with the visual directly in Outlook on the web. Access tooltips and explore data points. See filters and slicers applied at the time of sharing. View refresh information and data context. Navigate back to the source report when deeper analysis is needed. The result is a richer conversation around data, directly within Outlook. Sharing a live visual is simple: Open a Power BI report in the service. Select a visual to share, then select More options in the upper-right corner. Select Share. Select Link to this visual. A pop-up will appear asking if you want to use the current view (apply existing filters and slicers) or not. After selecting Create Link, the link is copied to your clipboard. Open Outlook on the web. This experience is not available in the Outlook desktop app. Create a new email. Paste the link in the email. The visual renders in loop after a few seconds! You can also share the same link in a Loop page, and the visual will also be rendered. By combining Power BI, Outlook, and Loop-powered experiences, we're bringing data closer to where collaboration happens. Live Visual Sharing can support more informed, data-driven conversations in Outlook. To learn more about Live Visual Sharing in Outlook, refer to the Share live Power BI visuals documentation. More control and context in Power BI Home (Generally Available) Power BI Home now provides additional context and personalization options to help users discover the right content more efficiently. With new freshness indicators and customizable list views, you can quickly identify the most relevant reports and semantic models, prioritize the metadata that matters most, and tailor the experience to fit your workflow. Find the right content more easily Discover the right content more easily with new Data Refreshed and Modified Date columns in Power BI Home. By bringing the freshness signals already familiar from workspace list views into Home, users can more easily spot recently updated reports and semantic models, distinguish between similar items, and quickly find the content that matters most. Whether you're checking for the latest business updates or deciding which refreshed report to open, you can navigate with greater confidence and less guesswork. Customize your Home list view Not every user, team, or organization relies on the same signals to discover content. With Column Options, you can show or hide metadata, drag and reorder columns to prioritize the information you use most, and reset to the default view at any time. Whether you rely on freshness signals, endorsements, sensitivity labels, or location information, you can tailor Home to match your workflow. To learn more, refer to the Power BI Home documentation. Set and reset column widths in table and matrix (Generally Available) You can now right-click a column header in a table or matrix and select Set column width to specify its width in pixels. This option helps report viewers improve readability and fit the information they need within the available space, without changing the report author’s design or affecting other viewers. Changes made by viewers last only for the current viewing session. To restore the width defined by the report author, right-click the column header and select Reset column width. To learn more, refer to the Adjust column width in Power BI tables documentation. Visual calculations for small multiples (Generally Available) You can now use visual calculations in visuals with small multiples. Create calculations such as moving averages and compare trends across categories without adding separate measures to the semantic model. In the visual matrix, the field in the small multiples field well appears on the ROWPAGES axis. You can select ROWPAGES by itself or combine them with ROWS and COLUMNS to control how the calculation moves through the visual. e: Create a moving average visual calculation for a line chart divided into small multiples by country. To learn more, refer to the Visual calculations overview documentation. Design ribbon for Power BI reports (Generally Available) The new Design ribbon brings report styling tools together in one location. Apply a built-in or organizational theme, open the Theme pane to customize report-wide defaults, or import and download theme files directly from the ribbon. You can also quickly open Page size, Page background, and Wallpaper settings for the current page. Two new built-in themes, Dark and Soft dark, use the new Versa color palette and styling that closely follows the Fluent 2 base theme. These themes let you design against a dark report canvas and return to the default light theme before publishing if needed. To learn more, refer to the Use the Design ribbon in Power BI documentation. Slider color options for Slicer visual (Generally Available) Match your slicer's slider to your report design. The Slicer visual's Between, Before, After, and Date picker settings include a Slider format section where you can turn the slider on or off, set the handle style, and choose colors for the handles, selected range, and track. Handle options adjust based on the Responsive layout toggle (General > Properties > Advanced options). When Responsive is on, handles appear as circles with Fill color, Border, and Size. When Responsive is off, handles appear as rectangular bars with Border, Width, and Height. To learn more, refer to the Slicer visual in Power BI documentation. Translytical task flows show user date function (UDF) information in button action (Generally Available) The formatting pane now displays the Workspace, Function set, and Data function assigned to a data function button. These read-only details help you confirm that the button references the intended function without reopening the OneLake catalog. This information is especially useful when you review or troubleshoot reports that use several functions or connect to functions across multiple workspaces. To learn how to configure data function buttons, refer to Create a data function button in Power BI. Format pane headers stay visible as you scroll (Generally Available) Expanded section headers now remain visible at the top of the Format visual pane as you scroll through their settings. This update helps you keep track of the section you’re editing, especially when a section contains many formatting options. For example, when you scroll through color settings, the Colors and Data colors headers remain visible for context. To learn more about formatting visuals, refer to Get started formatting report visualizations. Pick the axis initial scroll position (Generally Available) Categorical axes—like month, year, or product name—show a scrollbar when there are more categories than fit in the visual. Until now, the scrollbar always started at the beginning of the data, so the oldest values displayed first. This month, you can set Initial scroll position to End on the axis Layout section, and the visual loads with the most recent data in view. Users can still scroll back to see earlier history. The option is available on stacked, clustered, and 100% stacked bar and column charts; line, area, stacked area, and 100% stacked area charts; line and stacked column and line and clustered column charts; ribbon, waterfall, and scatter charts with a categorical X-axis. Initial scroll positionset toEndon the X-axis. The scrollbar is positioned on the right, showing the most recent months first. To learn more, refer to the Customize x-axis and y-axis properties documentation. Azure Maps geocoding improvements (Generally Available) Large reference layers now render correctly when you filter your report. Reference layers add geographic context using points, lines, and polygons, such as locations, routes, postal codes, sales territories, and administrative regions. The Azure Maps visual renders up to 30,000 reference-layer features at a time. Previously, when a reference layer exceeded this limit, features excluded from the initial view could remain missing even after a report filter narrowed the data to a smaller area. . . . Now, Azure Maps re-evaluates the reference-layer features whenever report filters are applied. The unfiltered view is still limited to 30,000 features, but a filtered view accurately displays the relevant features, up to that limit. On-object interaction improvements (Preview) Based on your feedback during preview, we're introducing the first in a series of on-object interaction improvements, designed to make quick edits to your visuals easier. This month we're refining how you add data and manage some visual elements. The Build flyout now opens only when you select Add data, helping you stay focused when dragging and dropping data directly onto a visual or into the Build pane. We’ve also separated Title and Subtitle in the Add to your visual flyout, so you can manage them independently. In addition, title and subtitle are now turned off by default for matrix, table, and card visuals in reports that use the latest base theme. New reports automatically use the latest base theme. For existing reports, select Update theme in the Theme pane to apply this change. You can turn either element back on when needed. to your visual menu. To learn more, refer to the using on-object interaction with visuals and updating to the latest base theme documentation. Support for Fabric Apps in Org apps (Coming Soon) Soon, users will be able to include generative Fabric Apps alongside the reports, dashboards, notebooks, and other content in your Org Apps. For example, imagine you want people to rate the content in an Org App and share feedback about what would make it more useful. You can create a Fabric App for that experience and add it directly to the Org App. Whether you’re helping people explore what could happen next, act on an insight, or complete a specialized workflow, Fabric Apps give you a flexible new way to build for your audience without sending them somewhere else. Together, organizational apps and Fabric Apps provide a way to bring enterprise reporting, content, and tailored experiences together for your organization. Learn more about org apps and item types supported, including Fabric Apps. Changes to Power BI subscription email delivery Power BI is updating the platform used to deliver subscription emails as part of ongoing improvements to delivery reliability and long-term service health. As eligible subscriptions move to the updated platform, customers may notice changes to the sender identity and email formatting. Each subscription transfers automatically when it moves. Customers do not need to recreate it or change its schedule or recipients. After selecting Subscribe for an item, an information icon identifies a subscription that has moved to the updated email delivery platform. Learn more about changes to Power BI subscription emails. Increased Export to Excel row limits (Coming Soon) Many users export Power BI data to Excel for further analysis. This month, we're increasing the maximum number of rows that can be exported from table visuals using Data with current layout from 150,000 to 500,000. This increase helps you analyze larger datasets in Excel with fewer export operations. s. To learn more about exporting data from Power BI visuals, refer to Export data from a Power BI visualization. Value selection in Narrative visual We are replacing the Q&A natural language experience for creating dynamic values in the Narrative visual and text box with a more deterministic UI. This is due to the ongoing deprecation of the Q&A tool, and also based on feedback that it was hard to deterministically pick exact values through the Q&A experience. Authors can now create and edit custom values through a direct, deterministic UI: select a column or measure, apply any valid summarization, and add any type of filter. Auto-generated values are unchanged, and values previously authored with Q&A remain available as read-only. To learn more, refer to the Create Smart Narrative documentation. Modeling Create and edit DirectQuery semantic models in the Power BI service DirectQuery semantic models can now be created and edited directly in the Power BI service, eliminating the need to download and use Power BI Desktop. Previously, building models that query data sources in real time meant downloading Desktop, authoring the model, and publishing it to the web. Now, web-based modeling gives your team the freedom to: Work from any device: Mac, Linux and Windows users can now author DirectQuery models straight from the browser. Collaborate in real time: Multiple team members can work on the same semantic model simultaneously. Ditch the desktop loop: Skip the download-edit-republish cycle and update your models instantly in the service. Capabilities on the web Start from scratch: Connect to a supported data source and choose DirectQuery mode. Figure: Power Query interface displaying data source connection settings, with DirectQuery selected as the data connectivity mode. Full modeling control: Open existing DirectQuery models to manage relationships, write DAX and tweak table/column properties. : Model View in Power BI Service showing tables, relationships, and the properties panel for editing a DirectQuery semantic model. Flexible storage modes: Easily switch table storage modes from DirectQuery to Import or Dual if needed. : Properties pane that allows user to switch table storage modes from DirectQuery to Import or Dual. Thanks to this new capability, composite models are also available on the web! Head to the Power BI service and open one of your DirectQuery semantic models or create a brand-new one. We'd love to hear what you think! Learn more in the DirectQuery documentation. Approximate Distinct Count comes to Import and Direct Lake (Preview) Distinct count is one of the most resource-intensive operations in any analytical engine. As data cardinality grows, so do CPU and memory usage, often slowing down your interactive reports. APPROXIMATEDISTINCTCOUNT already offered a faster alternative for DirectQuery sources such as Azure SQL, Synapse, BigQuery, Databricks, and Snowflake. This month we're extending support to Import, Direct Lake on OneLake, and Direct Lake on SQL semantic models. DAX = APPROXIMATEDISTINCTCOUNT(<columnName>) : Model View in Power BI Service showing tables, relationships, and the properties panel for editing a DirectQuery semantic model. How it works and when to use it Instead of scanning every single unique value, APPROXIMATEDISTINCTCOUNT calculates a highly accurate estimate of unique values with a standard error rate of ~1.6%. In return, you get faster query execution times. Best for high-cardinality columns: Use it on columns with millions of unique values (such as transaction IDs, user IDs, or timestamps) where an approximate result can reduce the resources required to calculate distinct values. Keep standard DISTINCTCOUNT for low-cardinality data: Stick to exact counts for smaller attributes like Country, Category, or Status, where standard calculation is already fast, or for financial metrics where absolute precision is required. Swap APPROXIMATEDISTINCTCOUNT into your heaviest measures today and let us know how much speed you gain on your models! Learn more in the APPROXIMATEDISTINCTCOUNT documentation. Accelerate text searches in your semantic models with String Indexing (Preview) String Indexing for Power BI semantic models is coming soon! If your reports rely heavily on text fields, such as product names, customer comments, or long descriptions, you’ve likely faced slow performance when users search or filter on large semantic models. String Indexing is designed to improve text-search performance in supported semantic models. Why String Indexing? Common text operations like DAX functions (SEARCH(), CONTAINSSTRING()) and advanced text slicers in reports traditionally scan data row-by-row. As datasets grow, these scans slow down query times. String Indexing builds specialized indexes so the engine can locate matching text more efficiently, which can improve report responsiveness for both Import and Direct Lake semantic models. Indexing modes You can configure String Indexing via model metadata to fit your workload needs: Auto Mode: Automatically builds the index on the first query execution for a qualifying column. The index does not persist in storage; it remains in memory only while the column is loaded and will be rebuilt on subsequent queries if evicted. Ideal for smaller models. Full Mode: Builds and persists the index during scheduled data refreshes. While refresh times and model size may slightly increase, users enjoy consistently fast query performance right out of the box. Get started String Indexing requires a semantic model running at 1707 compatibility level. During this initial preview phase, indexing can be configured through TMDL on the web or XMLA metadata. Desktop support and a dedicated UI experience within Power BI are planned for a future release to simplify setup. Check out the String Indexing documentation for setting up instructions, prerequisites, and guidance on choosing the right indexing mode for your models. We look forward to seeing how this feature helps make text-heavy reporting experiences faster and more responsive! Introducing Full-Text Search in Power BI Semantic Models (Preview) Full-Text Indexing for Power BI semantic models is coming soon! While traditional string searches look for exact character matches, Full-Text Indexing enables a more intelligent search experience based on the meaning of words and natural language concepts. This capability powers two new DAX functions, TEXTCONTAINS() and TEXTSIMILARITY(), making it easier than ever to analyze large volumes of textual data directly in your semantic models. Why Full-Text Indexing? Business data increasingly contains rich text content such as customer feedback, support tickets, product descriptions, incident summaries, and knowledge articles. Searching this data using traditional string matching often requires exact matches and can miss relevant results due to differences in spelling, word forms, or phrasing. Full-Text Indexing builds a specialized search index at the column level, allowing semantic models to understand words rather than just character sequences. This enables advanced search experiences such as: Finding related words through linguistic analysis (for example, run, running, and runs) Matching singular and plural variations automatically Ignoring common stop words such as "the" or "and" Supporting phrase-based searches Tolerating minor spelling mistakes through fuzzy matching Enabling semantic similarity search scenarios Getting Started FTS Indexing requires a semantic model running at 1708 compatibility level. During this initial preview phase, indexing can be configured through TMDL on the web or XMLA metadata. Desktop support and a dedicated UI experience within Power BI are planned for a future release to simplify setup. New DAX functions Full-Text Indexing is required for two new DAX functions: TEXTCONTAINS() Returns TRUE when a text value matches a search expression using full-text search semantics rather than simple substring matching. It supports multiple search modes, including keyword matching, phrase matching, and fuzzy matching for typo tolerance TextContains(<WithinTextColumn>, <SearchText>, [MatchMode]) Example: "Show me all reviews that mention spa." The review either matches or it doesn't. TEXTSIMILARITY() Returns a relevance score indicating how well the text in <WithinTextColumn> matches <SearchText>. It enables similarity-based text comparisons, helping identify records that are semantically related even when they do not contain the exact same words. TextSimilarity(<WithinTextColumn>, <SearchText> [, <MatchMode>]) Example: "Show me the 10 reviews most relevant to 'good breakfast'." Reviews mentioning both terms prominently rank higher than reviews that only partially match. Learn more about these two new DAX functions (TEXTCONTAINS, TEXTSIMILARITY) in our documentation. Full-Text Indexing is supported for both Import and Direct Lake models, with Full mode recommended (during Preview) for the best out-of-the-box performance and user experience. Refer to the Full Text Indexing documentation to learn more about prerequisites, configuration options, and setup instructions. Data connectivity New Power Query editor (Preview) Since May of this year, the modern Get Data experience has been available in preview inside Power BI Desktop; the feedback we've gathered from that rollout has directly shaped what comes next. Your reports of friction points, missing capabilities, and rough edges between the legacy and modern interfaces went straight into the backlog, and this month that work pays off: the full modern Power Query Editor experience moves into preview. It's the same trusted engine, wrapped in a completely refreshed editing surface — and we're still closing the remaining gaps with the legacy UI as we go. Let's explore six key reasons to enable and try this feature. Modern Get Data - The same connector browsing and search experience you already know from Fabric Dataflow Gen2 — now in Desktop. Searchable, categorized, and designed to make connectors easier to find and navigate than the legacy dialog. Try it: Connect to a new data source and browse the OneLake catalog. To learn more, refer to the Data sources in Power BI Desktop documentation. Dark theme, end to end - No more jarring light-mode panes breaking your flow. The legacy editor only partially supported dark theme. The modern editor applies it consistently across the preview grid, dialogs, and other parts of the editing surface. Folding visibility - Stop guessing whether your query folds. Step folding indicators mark each step as folding or not, and the new query plan shows exactly how the engine intends to evaluate your query — and where folding breaks down. Try it: Open the query plan on a query with several steps and find where folding stops. To learn more, refer to the Understanding Query Evaluation and Query Folding in Power Query documentation. Diagram view - See how your queries relate to each other visually, instead of reverse-engineering it from names. A gamechanger for files you've inherited or models with long reference/merge chains. Try it: Open a file with referenced queries and trace a merge visually. To learn more, refer to the Diagram view documentation. New capabilities, now on Desktop - Features that used to live only on the web are now here: rank column, cluster values, schema view, query script view, plus upgrades to transforms you already use daily. Try it: Add a rank column or cluster values on a messy text column. To learn more, refer to the Rank column and Use Schema View documentation. Export a Power Query template - Capture your solution at a point in time and take it with you. Export a template from Desktop and import it straight into a Dataflow Gen2 — queries, steps, and structure intact. Try it: Export a template from a multi-query file and import it into a Dataflow Gen2 to compare. Figure: Screenshot of the Export template experience in the Power Query editor. To learn more, refer to the Power Query template documentation. Ready to try it? Enabling the modern Power Query experience takes just a few seconds — no restart required: Open Power BI Desktop. Locate: File > Options and settings > Options. Under Preview features, check New Power Query experience. Open (or reopen) the Power Query Editor — you're now in the modern experience. Once you've had a chance to explore this new user interface and its experiences, we want to hear from you. Share your thoughts on the new Power Query experience in the Fabric Community — what's working, what's missing, and what would make it a compelling alternative from the legacy editor. Every comment helps shape what ships next. Microsoft Outlook Mail connector (Preview) We're introducing the new Microsoft Outlook Mail connector (Preview) for Power BI Desktop. Built on the Microsoft Graph API, it lets you read your mailbox data — folders, messages, and attachments — using an organizational account. This is the first in a set of new connectors we'll be releasing as a modern alternative to the Exchange Online connector. The initial preview is scoped to Mail in Power BI Desktop. Additional connectors for Calendar Events, Contacts, and Tasks are on the roadmap, along with Excel and Power BI service / Fabric Dataflows Gen2 support. Scheduled refresh isn't supported yet — this applies to both the Power BI service and refreshes routed through the on-premises data gateway. To try it, enable the Microsoft Outlook Mail preview feature under Options in Power BI Desktop, then open Get Data → Online Services → Microsoft Outlook Mail and sign in with your organizational account. To learn more, refer to the Microsoft Outlook Mail connector documentation. More flexible Data Privacy Firewall query patterns in Power BI Desktop (Preview) Power BI Desktop now allows Data Privacy Firewall partitions that reference other partitions also access data sources when the privacy levels of the involved data sources are compatible. Introduced in the July 2026 release and enabled by default, this preview setting aligns Power BI Desktop with Power Query Online, gateways, and the Power BI service. This change supports common query patterns in which one query returns a value that another query uses to build a web request, SQL statement, file path, or other data source request. These patterns can now run without refactoring queries or duplicating logic solely to satisfy the previous partition restriction. The Data Privacy Firewall remains active. Power Query continues to evaluate the privacy levels of the involved data sources before combining data, helping authors use more flexible query designs while preserving privacy protections. To learn more, refer to the Allow data privacy firewall partitions that reference other partitions to also access data sources documentation. Mobile Mobile page indicator (Generally Available) Know immediately which report pages are mobile-ready, without clicking through them. Each page that has a mobile layout shows an icon on its page tab in the footer, so you can spot your mobile coverage instantly. Pages without a mobile layout stay unmarked, making it easy to see what's left to optimize. The icon is coming soon to both Power BI Desktop and the Power BI service. In Power BI Desktop, it appears in both desktop layout view and mobile layout view, so you can track mobile coverage right where you work. In the Power BI service, it appears in mobile layout view. Page tabs marked with an icon to show which pages have a mobile layout. To learn how to build mobile-optimized reports, refer to the Mobile-optimized Power BI reports overview documentation. Contact details in Info drawer (Generally Available) We’re making it easier to connect with the right people from the Power BI Mobile app. You can now find contact details in the Info drawer for reports, apps, and dashboards, and email the relevant contact directly from your phone. To learn more about Power BI in mobile, refer to the What Are the Power BI Mobile Apps? documentation. Developers and APIs Power BI Projects (PBIP) (Generally Available) Since entering preview in 2023, PBIP has changed how developers build and manage Power BI content. It stores report and semantic model definitions as human-readable text files, turning a Power BI solution into a codebase. Developers can search these files, track changes in Git, review diffs in pull requests, automate updates with scripts and pipelines, and use AI agents to work with models and reports. Changes remain visible and reversible, which makes team development easier and safer. Recent improvements have made the workflow more practical. These include opening a project directly in Visual Studio Code and detecting external file changes in Power BI Desktop. New underlying formats capabilities: Tabular Model Definition Language (TMDL) for semantic models. Power BI Enhanced Report Format (PBIR) for reports. PBIR is now the default report format for PBIX and PBIP files. 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. To learn more, refer to the Power BI Projects documentation. New Purge Data option for Fabric APIs The new Purge Data option for Fabric APIs gives developers and administrators greater control over programmatic semantic model deployments. This capability allows users to explicitly approve deployments that may require removing existing data, helping unblock scenarios such as schema or data type changes that previously failed validation checks. Designed with both safety and automation in mind, purgeData keeps data protection safeguards enabled by default while giving developers the flexibility they need to manage complex deployment scenarios at scale. To learn more, refer to the Purge Data documentation. Other Data updated and sensitivity fields of Power BI report in Microsoft Teams preview card (Coming Soon) Power BI report previews in Microsoft Teams will soon surface the report's last refresh date and sensitivity label. These additional signals will help users quickly assess the freshness, relevance, and classification of a report before opening it. To learn how Power BI content is displayed in Microsoft Teams, refer to the Link preview cards in Microsoft Teams chats and channels documentation. Artifact info header for Power BI tabs in Microsoft Teams (Generally Available) Power BI content in Microsoft Teams tabs now includes the artifact info header, giving users the context they need without leaving their flow of work. Available for reports, apps, organizational apps, dashboards, and scorecards, the header brings key artifact information and actions together in one place. The new header displays the artifact name and sensitivity label and provides an entry point to artifact metadata. You can open the artifact in the Power BI personal app in Teams, open it in a browser, and refresh the tab. Tab owners can control their experience with the new Show info header setting on the tab settings page. The setting is 'on’ by default and can be turned off when a more compact tab experience is preferred. To learn how to add Power BI content as a tab in Microsoft Teams, refer to the Embed Power BI reports in Microsoft Teams documentation. Third-party custom visuals graphomate matrix The graphomate chartrix is a brand-new charting visual, and it offers various improvements over graphomate charts. First and probably most obvious is the new interactive user interface which enables users to define most elements of the chart layout directly on the visual. To keep this interface effective and uncluttered it focuses on the most important parameters of your chart. Users can now create combination charts which were simply not possible before. There are also new chart types like the variance waterfall, which automatically renders from one measure and two dimensions. Of course, the graphomate chartrix is IBCS® certified. With the latest release an XFL editor has been introduced, which helps you write scripts to format the chartrix far beyond the possibilities of the regular option set. XFL scripts also provide a way to conditionally format the visualization, dynamically changing properties in relation to the actual data. This new visual is provided free of charge to existing customers of the graphomate charts. All development efforts will be focused on making the graphomate chartrix the best IBCS® charting visual there is. Video Link: graphomate chartrix for Microsoft Power BI AppSource Link: graphomate chartrix Sample Report Link Drill Down Bubble PRO by ZoomCharts Drill Down Bubble PRO is a Power BI visual designed for exploring relationships, trends, and outliers across multiple dimensions of data. It adds richer exploration, selection, and analytical capabilities directly inside the visual. Key features include: Multi-level drill down to explore data hierarchies from a high-level overview to detailed insights. Pan and zoom controls to navigate dense datasets and focus on specific clusters or individual data points. Area and lasso selection to quickly isolate groups of bubbles and use selections for interactive analysis and filtering. Dynamic regression lines that recalculate as the displayed data changes, making trends and relationships easier to identify. Line and area thresholds to highlight targets, ranges, exceptions, or other important business benchmarks. Area shading to visually group and compare selected segments of data. 200+ formatting settings to tailor the chart to different reporting and branding needs. Figure: Drill Down Bubble PRO demonstrates multi-level drill down, panning, and zoom controls for exploring data from high-level patterns to detailed points. These capabilities make Drill Down Bubble PRO especially useful for customer segmentation, sales analysis, portfolio analysis, performance monitoring, and other scenarios where users need to actively explore patterns rather than simply display them. Get on Markeplace Commenting in view mode: introducing Zebra BI Tables+ & Charts+ It's month-end. A regional manager opens the report, spots a variance in their region, and knows exactly why it happened. But the report isn't theirs to edit, so the explanation ends up where it always does: a screenshot pasted into PowerPoint, emailed around, detached from the number that prompted it. That's the collaboration gap. Your report shows what happened; the why and the what next leave the report and get lost. Meet Zebra BI Tables+ and Charts+: two new visuals that keep everything existing Zebra BI Tables and Charts offer and add a built-in collaboration layer. Report viewers can add comments directly on a data point, in view mode, with no Power BI edit access. Each comment is anchored to the exact figure and stored in your own SharePoint file. Zebra BI stores nothing. ting files from SharePoint, and collaborate directly in view mode. So instead of screenshotting insights into PowerPoint or maintaining a custom workaround, the whole team explains and decides right on the report. Instead of slides or emails, the next steps can be defined directly on the spot. Stop losing good decisions to a missing context. Start making them on the report. Try Zebra BI Tables+ and Charts+ free for 14 days on AppSource. Network Graph by Powerviz The Powerviz Network Graph is a visual representation of Nodes and the relationships or connections between them. Its primary purpose is to illustrate how different entities are related or interact with one another. Key Features Line Options: Customise node links with different styles and colors. Node Shape: Choose from various shapes, icons, or upload your own. Labels: Display categories, values, and percentage labels. Data Colors: Access 30+ color palettes, including color-blind-safe options. Fill Patterns: Highlight nodes with patterns or upload your own. Conditional Formatting: Highlight outliers based on measures or category rules. Clusters: Group related nodes to reveal patterns. Relationships: Analyze node connections at a glance. Animation: Add pop-ups and highlight effects. Import/Export: Save and reuse themes easily. Toolbars: Gain advanced control over the graph. Other features included are Sorting, Ranking, Lasso / Reverse Lasso, Grid View and more. Business Use Cases Customer Relationship Management (CRM), Supply Chain Optimization, Social Network Analysis (SNA). Resources Try Network Graph visual for FREE from AppSource Check out all features of the visual Step-by-step instructions YouTube Video Learn more about visuals Follow Powerviz on LinkedIn Hierarchical Bar Chart New, exciting changes were made to Hierarchical Bar Chart. Bubble Matrix for stacks - The visual now provides the option to show stacks as bubbles below the bars. lues into bar-and-bubble matrix layouts for clearer comparison of category totals and component contributions. Display Stacks as Clustered Columns - From the format pane settings, you can choose to display stacks as clustered column chart. Filter using legends - You can click on a stack legend to filter out other stacks. On mouse hover relevant stack gets highlighted. Stack ribbons and %age of total - You can now choose to connect stacks with ribbon and show percentage of total for each stack. Resources Check out the Hierarchical Bar Chart Update YouTube video. Download the visual from APPSOURCE Download a sample pbix file For more information refer to Hierarchical Bar Chart or contact us. That’s a wrap! The European Microsoft Fabric + SQL Community Conference is just getting started, and there's never been a more exciting time to be part of the Power BI community. We hope these updates help you discover new capabilities, streamline your workflows, and get even more value from your data. Be sure to explore the links throughout this post for deeper technical guidance and follow along with the latest announcements and sessions from Barcelona.Introducing the data engineering agent for Fabric (Project Osmos)
AI coding assistants are excellent at turning natural-language transformation instructions into code. For data engineers, the harder question is whether that code produces results they can trust. You can describe transformation logic in natural language, and an AI coding assistant will happily generate hundreds of lines of Spark code. Impressive. But a Spark job can execute successfully and still produce completely wrong results. The logic may work for a subset of the data but fail when run against the full dataset. The pipeline's logic and validation checks may not account for future changes to the data or schema. Incremental loads can cause duplicates. Revenue numbers might look perfectly reasonable until someone reconciles them against the source. Anyone who's spent time working with real-world enterprise data knows that automating data engineering with AI comes with a unique set of challenges. Data engineering involves more than code: it requires working with large datasets and retaining context about their schemas, relationships, and business rules over time. Quality & governance matter a lot. The cost of errors or data loss is very high. Data engineering work typically spans team-based, collaborative, long-running, milestone-driven projects; not a single conversation between one person and a coding agent. What if specialized, data-aware agents could help teams investigate these problems, implement solutions, and validate the results—not just generate code? Following Microsoft's acquisition of Osmos in January 2026, we've been working on bringing autonomous data engineering capabilities as a native experience in Microsoft Fabric. The idea behind Project Osmos is simple: instead of asking AI to help you write data engineering code, you define a complete problem and the outcome you expect. It has the autonomy to use the tools and an orchestration framework to build, validate, troubleshoot, and refine its implementation, making decisions within the boundaries you set and working towards a high quality output. The task's code execution and data operations run within your selected Fabric lakehouse. Real-world data engineering is rarely straightforward Let's take an insurance use case to illustrate the complexity. Imagine an insurance company consolidating policy and claims data from enterprise systems such as SAP, Guidewire, and Oracle, a common challenge in an industry that grows through acquisitions. The project also needs to ingest external datasets and user-generated data. The objective is to build a consolidated view of a claim, including its policy, coverage, and financial history. Sounds reasonable, until you start looking at the data. The same person, vehicle, or incident could have different IDs across systems. The same ID could also refer to completely different incidents across systems. Reconciling payments and ledger entries needs to be done with extreme caution. Discrepancies could result in overpayments and wrong forecasts. You need to investigate the data, understand its relationships, establish the business rules, and validate that the resulting view is consistent with those rules. Getting it wrong can have serious downstream implications. Quality and accuracy matter more than speed. In this example, a team would ask the Project Osmos to investigate relationships across the datasets, implement the agreed business rules, and validate the consolidated claims view. The agent would autonomously work through the problem and regularly report updates to the user. Users can review the evidence and validate results before accepting the output. A purpose-built orchestrator Under the hood, Project Osmos uses a purpose-built orchestrator that coordinates multiple specialized agents to tackle the problem. Each agent has a well-defined role from code generation to validation to evaluation. The agents interpret instructions, analyze the data, evaluate implementation approaches, generate code in isolated sandboxes, and verify the outputs. They use execution results to guide further iterations, with the aim of meeting your validation and quality requirements. Your team reviews the evidence to determine whether the results are acceptable. Should the agent need additional information or get blocked, or if the agent determines that the task cannot be accomplished within the constraints, the user will be elicited for input and guidance, and task will pause waiting user input. Project Osmos isn't merely simulating what the code might do. It submits PySpark code through the Fabric Livy API for execution in Fabric Spark sessions, so it can test the implementation and validate actual results. Whether changes use copies, staging tables, or in-place updates depends on the write pattern you select. The goal isn't to remove humans from the process. It's to let the agents take on the investigation and implementation work, while allowing people to guide the outcome, resolve ambiguities, and review the evidence. Autonomy, with the right guardrails Of course, giving an AI agent the autonomy to execute code and modify data raises the most important question: How do you make sure it doesn't do something it shouldn't? Before a task starts, you review the operating guardrails recommended by Project Osmos to define the scope and impact of its work. You specify which resources it can use, where outputs should go, whether changes should be staged, how reruns should behave, and whether schema changes are permitted. Project Osmos also recommends a reasoning effort setting, which users can adjust. More extensive reasoning can be useful for complicated problems, but it also comes at the cost of additional time and compute. These instructions guide the agent's work, but they aren't a substitute for proper access controls. Agents operate within the user's permissions, and Fabric and OneLake permissions continue to enforce what the agent is authorized to access. And while Osmos can carry out extensive validation, your team ultimately determines whether the results meet the business requirements. A persistent task that lives in Fabric Real-world data engineering problems rarely get completed in a single sitting. Requirements change. New data arrives. Business rules get clarified. Often implementations that worked perfectly the previous month need to be revisited. With Project Osmos, work is organized into persistent tasks hosted in Microsoft Fabric. Each task represents an ongoing data engineering project, rather than a single chat exchange. Once created, a task runs independently in Fabric, consuming Fabric capacity units (CUs). You can start it from GitHub Copilot CLI, Claude Code, or Codex using the Skills for Fabric plug-in, then disconnect without stopping the work. You don't need to keep your terminal open, maintain an active chat session, or even keep your laptop running. The task maintains its state, context, execution history, and outputs. You can monitor its progress in your Fabric Lakehouse, inspect what the agents are doing, review errors, and examine the outputs. The task is reusable and can evolve with additional instructions. You can interrupt it to provide additional context, clarify a requirement, or pause and resume the work through your local coding agent. Say you've built a data pipeline, and a new data drop arrives the following week. You don't have to start another task and explain the entire project all over again. You can reference the existing task and ask Project Osmos to process the new data. It has the context of the previous work and can decide whether to reuse the existing implementation or update it to account for the new data. Collaborative by design Complex data engineering projects are rarely a one-person job. They involve a team of data engineers, data architects, business analysts, data stewards, and domain experts. And each of these people brings something different to the table. Getting the details right often requires conversations between people who understand different aspects of the problem. Project Osmos allows authorized teammates to work with the same task, inspect what's been done, and provide additional guidance. So instead of passing notebooks back and forth or having different people initiate separate AI conversations with slightly different assumptions, the team can collaborate on the same ongoing work. The task retains the context as requirements evolve, making it possible for someone joining later to understand what has already been attempted and contribute to the next steps. What about the data engineering work you've already done? So far, we've talked mostly about building new data engineering projects. But a significant amount of engineering effort also goes into maintaining and improving what's already there. This is also work you can delegate to Project Osmos. For instance, you could ask Project Osmos to convert an existing pipeline to an incremental load and validate its rerun behavior. You could also ask it to optimize Spark code in existing notebooks, improve the performance of Delta tables in your lakehouse, or assist with migrating data engineering workloads from another platform to Fabric while preserving business rules and validating the results. Give Project Osmos a try! Getting started with Project Osmos is straightforward. Install the Skills for Fabric plugin in GitHub Copilot CLI, Claude Code, or Codex, sign in to Azure, and ask your coding agent to create a Project Osmos task. Point it to your Fabric workspace and Lakehouse, describe the outcome you want, and specify how you'd like the results to be validated. You'll need Contributor or higher workspace permissions, the required tenant settings, and an eligible Fabric capacity. Trial capacities are not supported. Review the quickstart for prerequisites and preview limitations before you begin. We've put together a few resources to help you get started: Project Osmos Quickstart: A step-by-step walkthrough, including a sample medallion architecture project you can run in an empty Lakehouse. Project Osmos Overview: A deeper look at how tasks work and what the agent can do. Best Practices: Guidance for defining project outcomes, setting boundaries, and specifying validation requirements. Have a messy dataset, a pipeline that needs fixing, or a migration you've been putting off? Give Project Osmos a try and see what it can do!Govern your Microsoft Fabric data estate from one place with the OneLake catalog and Govern Skills (Generally Available)
The Govern experience in the OneLake catalog is now generally available, providing centralized governance capabilities for tenants, domains, capacities, workspaces, tags, and policies in Microsoft Fabric. Govern Skills are also now available in the Skills for Fabric collection, giving AI agents the knowledge they need to help admins and data owners audit, investigate, and remediate governance issues at scale. Both were announced at FabCon Europe 2026. Together, they help you monitor, manage, and act on your data estate from one place, whether you work in the Fabric portal or with an AI agent. Why it matters As AI makes it easier for more people to build with data, data estates grow in size and complexity. More workspaces, items, and access requests create more decisions about ownership, organization, and protection. Centralized, proactive governance helps admins support new projects without turning every decision into a manual review or losing visibility into how data is used. Governance can't rely on periodic, manual audits anymore. It needs to be proactive, centralized, and able to scale with the estate. Capabilities with Govern experience and Govern Skills: Manage governance in one place instead of moving between separate admin settings, reports, and management pages. Understand your governance posture through estate-wide signals, gaps, and recommended next steps. Use AI-powered assistance to audit and remediate governance issues across many domains, workspaces, and items at once. Manage governance centrally in the OneLake catalog The Govern experience in the OneLake catalog brings management and administration together, so the people responsible for the data estate can see its state and act on it without leaving the catalog. The Govern tab provides two ways to focus your work: Tenant view - All data in Fabric page: assess the health of your entire data estate and use recommended actions to identify where to focus first. Owner view - My items page: focus on governance issues for items you can update. From either view, select View more to open the full insights report, explore governance trends, and see where gaps are concentrated. The following image shows the Govern experience in the OneLake catalog, with estate-wide insights, management capabilities, and recommended actions on a single page. The Govern experience also brings key management areas into the same place: Domain management: organize the estate around the way your business operates, manage domain ownership, and delegate responsibility to the right teams. Capacities: View and manage all capacities for your tenant. Workspaces: View and manage all workspaces in your tenant Policies Center (Preview): Configure, activate, and monitor governance policies across tenants, capacities, domains, and workspaces from one place. Policies in Fabric support attribute-based controls; for example, you can allow only a data engineering security group to create Dataflow Gen2 items in approved workspaces. Tags: View and manage existing tags or define data management tags for your organization. The following GIF shows how you move from the governance insights to managing domains, workspaces, and policies, all within the Govern experience. Know where to act with governance insights Visibility is only part of governance. The Govern experience surfaces estate-wide signals, gaps, and recommended next steps, such as establishing sources of truth with endorsements or increasing Sensitivity label coverage. Instead of starting from a blank page, you start from a prioritized view of what matters most in your estate. Customize the governance report The insights behind View more are also available as a report in the Admin monitoring workspace, which every Fabric administrator can access. Its semantic model is curated for customization, so you can build your own reports on top of it, for example to add alerts, extra dimensions, or longer retention. The report is available to all tenants regardless of license or size, and admins can share it with others. Data owners get their own version of the report in My workspace. Govern with AI using Govern Skills Govern Skills are a collection of AI skills, available in Skills for Fabric, that help users assess, understand, and improve the governance posture of their Fabric environment. Rather than navigating reports, APIs, and admin experiences by hand, you can ask governance questions, investigate issues, and receive guided remediation recommendations through a conversation with an AI agent, such as GitHub Copilot in Visual Studio Code. Built on the OneLake catalog governance pillars Govern Skills align to the core governance pillars of the OneLake catalog: Health: keep the estate clean, active, and well organized. Protect: keep the estate safe and compliant. Trust: keep data understandable, discoverable, and trusted. What's included The initial set of skills covers all three pillars, with the broadest coverage in Health: Health: inventory the tenant, assess governance health, and detect empty domains, unassigned workspaces, inactive or unused items, and missing governance metadata. Health skills can also use Fabric admin APIs for actions such as bulk workspace-to-domain assignment and domain admin management. Protect: check sensitivity label coverage and domain default labels. Trust: check description, tag, and endorsement coverage, and identify which items to curate first. For every pillar, Govern Skills provide prioritized recommendations and remediation guidance. If a signal cannot be measured, the skill identifies the limitation instead of returning a potentially misleading result and directs you to the Govern tab or Microsoft Purview for review. A consistent operating model Each Govern Skill follows the same workflow: audit, explain, recommend, and remediate. Rather than returning raw data or a list of issues, each response includes the finding, supporting evidence, impact, priority, and recommended action. The skills also include guardrails. They use supported Fabric admin APIs, don't invent missing details, surface uncertainty, and request your confirmation before sensitive or destructive actions. The core idea: governance expertise, not just API access Giving an agent access to the Fabric admin APIs provides technical capabilities. Govern Skills add governance expertise, guardrails, and operating procedures to those capabilities. Govern skills provide an AI agent with the context a Fabric admin would use: the jobs to be done, the issues to investigate, the evidence to collect, the criteria for prioritizing findings, and the steps for remediation. This context reduces the agent’s need to infer governance practices, leading to more consistent results and a more focused path to completing each task. In short, Govern Skills turn governance expertise from something each model has to reconstruct into reusable knowledge provided by Fabric. Example: run a governance health check on your tenant The following example demonstrates how a Fabric admin might use Govern Skills to review the health of their estate: Install Fabric Skills for GitHub Copilot CLI or other compatible AI coding tools. Ask the agent to run a governance health check for your tenant. Review the findings. The agent inventories your tenant and highlights issues such as empty domains, workspaces that aren't assigned to a domain, and inactive items, along with the evidence, impact, and priority of each finding. Choose a recommended action, for example assigning a set of unassigned workspaces to the right domain in bulk. The agent asks for your confirmation before it makes the change. Return to the Govern tab in the OneLake catalog to track how your governance posture improves over time. The following images show GitHub Copilot in Visual Studio Code responding to a request to run a governance health check on a Fabric tenant, using fabric Govern skills. The same health check also reviews the Protect and Trust pillars. The following image shows the Protect and Trust findings for a tenant, followed by a prioritized sequence that explains which actions to take first and why. A few best practices as you get started: Start with a tenant-wide health assessment to establish a baseline before you remediate. Address the highest-priority findings first, such as unassigned workspaces and empty domains, so that domain-based governance applies across more of your estate. Review each proposed change before you confirm it, especially bulk actions. What's next This release is the first step toward making governance AI-first and accessible to both people and agents. Future investments will expand Protect and Trust scenarios, add remediation capabilities, and integrate governance more deeply into Fabric AI experiences. Next steps For more information on the Govern experience - Govern your data in the OneLake catalog. To get started with Govern Skills - Skills for Fabric overview. For more information on policies - Policies in Fabric. For all FabCon announcements - FabCon and SQLCon 2026 in Barcelona: Building the data foundation for Microsoft Copilot and Agents. Share which governance scenarios you'd like Govern Skills to cover next in the comments on this post or on Fabric Ideas.DAG Connect Nov 2026- Things I'm Doing to Stay Relevant as a PBI in an AI World
AI is changing how we build Power BI reports. It can write DAX, suggest visuals, summarize insights, and even create report pages within seconds. As a Power BI developer, I’ll be honest.. this can feel exciting and a little scary at the same time. In this session, I’ll share the practical things I’m doing to grow with AI instead of competing against it. This is not a session about predicting whether AI will replace developers. It is an honest and friendly conversation about how our role is evolving and how we can prepare for what comes next. You’ll leave with practical ideas you can start applying immediately and, hopefully, a little more confidence about your future as a Power BI developer. Speaker :Santhana Lakshmi Ponnurasan , Power BI DataViz World Championships 2025 and 2026 Finalist | Microsoft MVP Organizers: Rajendra Ongole, Koundinya Lanka,Shashi,Navya K, Co-Team Charitha Reddy, Meghana,Sreekar
Connect coding agents to Fabric notebooks with FNTK (Preview)
The Fabric Notebook Toolkit (Preview) helps developers shorten the feedback loop between writing code and validating it with live notebook results. It connects coding agents to Microsoft Fabric notebooks through a command-line interface (CLI) that runs in the terminal alongside tools such as Claude Code, GitHub Copilot CLI, and Codex. An agent can inspect the current code, run one cell, review its output, make a targeted change, and run it again without resubmitting the entire notebook—much as a human programmer working interactively. A coding agent uses FNTK from the terminal to interact with a Microsoft Fabric notebook. From code suggestions to notebook execution A coding agent can help write a transformation, but a proposed change is not the same as a working result. When execution happens elsewhere, developers must move code into a notebook, run it, and bring the results or errors back to the agent. FNTK connects those steps so the agent can work with a running Fabric notebook and its execution results instead of relying only on code shared in the conversation. Interactive development and debugging are the central use cases. You can ask an agent to inspect a notebook, investigate a failing transformation, or validate a change using actual execution output. The notebook remains in Fabric and uses Fabric capacity for distributed data processing, while FNTK provides the terminal commands the agent uses to interact with the live session. Run, inspect, and refine one cell at a time FNTK keeps the notebook session active while the agent works, so developers can explore and refine code without repeatedly restarting the workflow. For example, an agent can inspect a schema, adjust a conversion, check several records, and return to the same transformation as new information emerges. This mirrors the iterative way developers work in notebooks and keeps each change grounded in actual execution results. The toolkit brings the operations needed for that workflow into the terminal. An agent can discover the target notebook, read and update its cells, execute code, and retrieve results. Instead of treating every adjustment as a new whole-notebook submission, it can focus on the code being investigated. Find workspaces, notebooks, and lakehouses to establish the context for the task. Read, add, or update notebook cells as the work develops. Execute a cell's code and inspect its output before deciding what to change. Reuse the live session to refine and rerun the relevant code. For example, a transformation may fail because an input column has a different type than expected. An agent can inspect the schema, revise the affected code, and rerun it in the existing session to check the result. This avoids restarting the entire notebook merely to test that change and keeps the investigation focused on the failing operation. Give the agent the evidence to debug FNTK gives the agent the execution evidence needed to determine the source of a failure. It exposes cell output and Spark job state, along with driver logs, executor logs, and logs from Livy, the service that manages Spark sessions. With this context, the agent can investigate the likely source of a problem instead of relying on source code alone. The CLI can return structured JavaScript Object Notation (JSON) responses that identify results, errors, and suggested next actions. An agent can use this information to determine whether execution has finished, retrieve diagnostics, and choose what to investigate next. This shortens the feedback loop between writing code and understanding its behavior while reducing the need to copy outputs and logs manually between tools. FNTK in practice at Microsoft For our data engineering team at Microsoft, FNTK filled a gap between generating code and working with real data. Previously, we used an AI assistant to draft code locally, then copied it into Fabric notebooks to run it and brought the results back into the conversation. With FNTK, we can stay in GitHub Copilot, execute code in Fabric, and use the results to iterate. We started by installing the toolkit, connecting with the appropriate permissions, and asking questions about a data source. Exploratory work quickly became a valuable use case. We could ask for row counts, distributions, or samples, then follow up based on the actual results. From there, we built shared workflow skills and a developer plugin around FNTK, incorporated our engineering standards, and connected the plugin to Azure DevOps through the Model Context Protocol (MCP) to bring user-story context into the same workflow. In practice, this has sped up development and helped us work across multiple user stories, with more of our time spent reviewing changes rather than writing every line ourselves. Copilot handles the reasoning; FNTK provides the connection to live notebook execution. We can still inspect the runs in Fabric, and our existing source-control, quality, and human-review practices remain in place. In practice, this workflow has helped us work across multiple user stories and spend more time reviewing changes instead of writing every line ourselves. Copilot handles the reasoning; FNTK provides the connection to live notebook execution. We can still inspect the runs in Fabric, and our existing source-control, quality, and human-review practices remain in place. Figure: FNTK connects an AI developer agent to live Fabric notebook execution, alongside MCP tools for context and delivery tools for human-reviewed changes. Capabilities with FNTK commands Command or group Usage fntk help Show command guidance and examples. fntk init Install or update coding agent workflow skills. fntk discover Find workspaces, notebooks, lakehouses, and identity information. fntk notebook Manage notebooks, sessions, and default lakehouses. fntk cell Read, edit, and organize notebook cells. fntk code Execute code, retrieve results, and cancel runs. fntk monitor Inspect Spark progress, diagnostics, and resource usage. fntk logs Download driver, executor, Livy, and cell logs. fntk resource Manage notebook resource files and folders. fntk skill Read built-in coding and debugging guidance. fntk server Manage the local FNTK server. Get started To try the workflow, install the toolkit and connect your coding agent by following the FNTK installation and quick-start documentation, including fntk init for workflow skill setup. Then choose an existing development notebook and a focused task, such as inspecting one transformation. Run the code, review the output, and use the result to guide your next change.
Retiring the Fabric data agent custom tool in Microsoft Foundry: move to Fabric IQ
In March 2025, we announced that agents in Microsoft Foundry could use Fabric data agents to answer questions about data in Microsoft Fabric. In the earlier integration, you connected to a data agent by adding it as a custom tool. For details, refer to Consume a data agent in Microsoft Foundry. What’s changing We are deprecating the earlier custom-connection integration. Starting 11/6/2026, that integration will no longer be supported. If your Foundry agents use a Fabric data agent as a custom tool, you must migrate them before that date. The replacement integration (Preview) uses the Fabric IQ (OneLake Catalog) tool, which makes your data available to agents and applications. In the Foundry portal, you can use this tool to add a published Fabric data agent to a Foundry agent. For configuration steps, refer to Consume Fabric data agent from Microsoft Foundry via Fabric IQ. Capabilities with the new integration The replacement integration changes how you find and connect Fabric data agents: Select accessible published data agents from the catalog. In the Foundry portal, the OneLake Catalog shows the published data agents available to you by name, so the connection workflow does not require you to enter a workspace ID or an artifact ID. Connect more than one data agent. A Foundry agent can call multiple data agents, such as separate agents for sales data and customer support data. The earlier custom-connection integration only supported one Fabric data agent. Connect through a Model Context Protocol (MCP) endpoint. MCP is an open standard for how an agent calls a tool. Each selected Fabric data agent is presented to the Foundry agent as a tool through a MCP endpoint. Coming soon, when a data agent answers a question, its response will also include the query it used to generate the answer. What you need to do If any of your Foundry agents use a Fabric data agent as a custom tool, complete the following steps before 11/6/2026: In Microsoft Fabric, make sure to publish your Fabric data agent with a rich and detailed description. Open the agent in the Microsoft Foundry portal. In the Tools section, add the Fabric IQ (OneLake Catalog) tool, and then select your data agent from the catalog. Update the Foundry agent's instructions. Name each data agent and describe the types of questions it should handle. Remove the earlier custom tool and then save the Foundry agent. Test the Foundry agent with representative questions for each connected data agent. If you connect data agents in code, update your implementation to use the Fabric IQ tool. For samples in Python, C#, and JavaScript, refer to Connect agents to Microsoft Fabric with Fabric IQ. Guides to migrate from Fabric Data Agent to Fabric IQ tool Option 1: Migrate in Foundry Portal Navigate to Foundry Portal and open the Tools Catalog, select “Fabric IQ (OneLake Catalog)”. . In the OneLake catalog, filter by “Data agent” type. Select any data agent you want to use, and you are ready to go. Provide instructions to your Foundry agent about how to use Fabric data agent. . Option 2: Migrate in SDK/Rest API After you publish the Fabric data agent, open its Settings and go to the Model Context Protocol tab. This tab shows: Data agent MCP server name MCP server URL (copy this value; you can use in every client) Data agent MCP tool name MCP server tool description Once you get the data agent MCP server URL, proceed with the following steps: Step 1: Get an access token az account get-access-token --resource https://management.azure.com --query accessToken -o tsv Step 2: Create a Foundry connection to the Fabric data agent A connection stores the Fabric data agent endpoint and how to sign in to it. Your agent uses this connection to reach Fabric. Replace each value in angle brackets, then run the request: curl --request PUT \ --url 'https://management.azure.com/subscriptions/<azure-subscription-id>/resourceGroups/<resource-group>/providers/Microsoft.CognitiveServices/accounts/<foundry-account-name>/projects/<foundry-project-name>/connections/<connection-name>?api-version=2025-04-01-preview' \ --header 'Authorization: Bearer <access-token-from-step-1>' \ --header 'Content-Type: application/json' \ --data '{ "type": "Microsoft.MachineLearningServices/workspaces/connections", "properties": { "authType": "UserEntraToken", "group": "AzureAI", "category": "RemoteTool", "target": "<fabric-data-agent-mcp-endpoint>", "audience": "https://api.fabric.microsoft.com/.default", "isSharedToAll": true, "sharedUserList": [], "UseCustomConnector": false, "Credentials": { "ClientId": "fake" }, "metadata": { "ApiType": "Azure" } } }' Main field capabilities: Field Value What it means authType UserEntraToken The agent calls Fabric as the signed in user, with that user's Microsoft Entra token. category RemoteTool The connection points to a remote tool, in this case an MCP endpoint. target Your MCP endpoint The address of your Fabric data agent. audience https://api.fabric.microsoft.com/.default The token is issued for the Fabric API. isSharedToAll true Every user in the Foundry project can use the connection. Credentials.ClientId fake Keep this value as shown. The response includes an id field. This is the full resource ID of the connection. Save it for step 4. It looks like this: /subscriptions/<azure-subscription-id>/resourceGroups/<resource-group>/providers/Microsoft.CognitiveServices/accounts/<foundry-account-name>/projects/<foundry-project-name>/connections/<connection-name> Step 3: Install the Python packages pip install "azure-ai-projects>=2.2.0" azure-identity Step 4: Set the environment variables Variable Where to find it FOUNDRY_PROJECT_ENDPOINT The Overview page of your Foundry project. FOUNDRY_MODEL_NAME The deployment name of your model in your Foundry project. FABRIC_IQ_PROJECT_CONNECTION_ID The full connection resource ID from step 2. Use the full ID, not only the connection name. export FOUNDRY_PROJECT_ENDPOINT="<your-project-endpoint>" export FOUNDRY_MODEL_NAME="<your-model-deployment-name>" export FABRIC_IQ_PROJECT_CONNECTION_ID="<your-connection-resource-id>" Step 5: Create an agent that uses the Fabric IQ tool and ask it a question This script creates a prompt agent with the Fabric IQ tool, sends it one question, prints the answer, and then deletes the agent. import os from azure.identity import DefaultAzureCredential from azure.ai.projects import AIProjectClient from azure.ai.projects.models import PromptAgentDefinition, FabricIQPreviewTool endpoint = os.environ["FOUNDRY_PROJECT_ENDPOINT"] with ( # Signs in with your Azure command line interface account (from az login). DefaultAzureCredential() as credential, AIProjectClient(endpoint=endpoint, credential=credential) as project_client, project_client.get_openai_client() as openai_client, ): # 1. Define the Fabric IQ tool. It uses the connection you created in step 2. fabric_iq_tool = FabricIQPreviewTool( project_connection_id=os.environ["FABRIC_IQ_PROJECT_CONNECTION_ID"], require_approval="never", # The agent calls the tool without asking you first. ) # 2. Create the agent with your model, its instructions, and the Fabric IQ tool. agent = project_client.agents.create_version( agent_name="MyAgent", definition=PromptAgentDefinition( model=os.environ["FOUNDRY_MODEL_NAME"], instructions="Use the available Fabric IQ tools to answer questions and perform tasks.", tools=[fabric_iq_tool], ), ) print(f"Agent created (id: {agent.id}, name: {agent.name}, version: {agent.version})") # 3. Ask the agent a question. Change this to a question about your own data. user_input = "Which customers placed orders above $10,000 last quarter?" response = openai_client.responses.create( input=user_input, extra_body={"agent_reference": {"name": agent.name, "type": "agent_reference"}}, ) print(f"Agent response: {response.output_text}") # 4. Delete this agent version so unused versions don't build up in your project. project_client.agents.delete_version(agent_name=agent.name, agent_version=agent.version) print("Agent deleted") Finally, run the script. python fabric_iq_agent.py Learn more and share feedback We recommend that you move to Fabric IQ as soon as you can. We look forward to improving this integration, and we’d like to hear your feedback. Join the community discussion to share feedback and report issues. For more information, see these articles: Consume Fabric data agent from Microsoft Foundry via Fabric IQ Connect agents to Microsoft Fabric with Fabric IQ
