In Short: What Is Rayfin?
Rayfin is Microsoft's open-source SDK and command-line tool, announced at Build 2026, that lets developers - and, just as deliberately, AI coding agents - describe what they want to build and get an enterprise-grade application backend generated directly into the application code: a database, authentication, and the supporting services around them. The backend deploys to Microsoft Fabric, which means the application's operational data lands in OneLake from day one.
In Microsoft's own words, Rayfin lets teams move "from prototype to production without managing infrastructure". The part that matters for data leaders is the destination: applications built this way put their data inside the governed Fabric estate rather than in yet another disconnected database.
If you are weighing up what this means for your own estate, our Microsoft Fabric consulting services and implementation cost guides for the UK, US and South Africa are a practical starting point.
The Problem Rayfin Solves
There has always been a gap between writing application code and standing up the backend it needs. Provisioning a database, wiring authentication, configuring hosting, and connecting it all securely is undifferentiated work that every project repeats - and it is exactly the kind of work that stalls a prototype on its way to production.
AI coding assistants made the gap more visible, not smaller. An agent can generate excellent application code in minutes and then hit a wall at "now deploy it somewhere real, with a real database and real auth". Rayfin exists to close that wall: the same natural-language description that produces the code can now produce the backend, generated into the codebase rather than configured in a portal.
What Rayfin Actually Does
Three things define it:
Backend generation from a description. A developer or a coding agent describes the application, and Rayfin generates the backend into the application code - schema, database, authentication, and supporting services - rather than returning a diagram or a checklist. It is an SDK and CLI, so it fits code-first workflows and CI/CD rather than portal clicking.
Deployment to Microsoft Fabric. The generated backend runs on Fabric, and the application's data lives in OneLake. There is no separate infrastructure estate to manage; capacity, storage and governance come from the Fabric platform you already run.
Open source, agent-first. Rayfin is published as an open-source project and designed to be driven by coding agents as much as by people. In Microsoft's Build 2026 framing, this is part of making Fabric "the place where AI agents build and run software".
Why the OneLake Destination Is the Strategic Part
For organisations already running Fabric, the significance of Rayfin is not faster prototyping. It is where the data ends up.
Every conventional app project creates a new operational database that the analytics team later has to integrate - another pipeline, another identity mapping, another source of drift. An application whose backend was generated onto Fabric starts life inside the estate: its tables sit in OneLake next to your conformed data, Purview governance applies, and Power BI, Fabric IQ and Fabric Data Agents can reason over the application's data without an integration project.
That collapses the oldest divide in enterprise data: the one between the systems that run the business and the platform that analyses it.
What Rayfin Is Not
Worth stating plainly, because the name travels faster than the definition.
Rayfin is not a natural-language query layer, not a BI feature, and not an intelligence layer over your existing data - that role belongs to Fabric IQ and the wider Microsoft IQ family. Rayfin builds application backends. If your goal is asking questions of governed data in plain English, you want Fabric IQ and Data Agents; if your goal is shipping an application whose backend runs on Fabric, you want Rayfin.
It is also new. Generated backends still need the same engineering judgement as hand-built ones - data modelling, security review, an operating model - and teams should treat early output as a strong scaffold rather than a finished production system.
What It Means for Your Teams
Application developers
The backend stops being a separate project. Describe, generate, iterate in code, deploy to Fabric - with authentication and the database included rather than bolted on. For internal tools and line-of-business applications, the distance from idea to production shortens materially.
Data teams
Operational data from Rayfin-built applications arrives in OneLake by default. That is a gift and a responsibility: the analytics estate inherits new sources without pipeline work, and the governance model has to be ready for application workloads, not just analytical ones.
Platform owners
Fabric capacity now serves applications as well as analytics. Capacity planning, workspace structure and environment strategy need to account for backend workloads - the same discipline, applied to a wider surface.
Getting Your Estate Rayfin-Ready
Rayfin lands best on an estate that is already well-governed. An organisation with a clean medallion architecture, sensible workspace boundaries and working capacity management can absorb application backends without drama. An organisation still untangling its analytical estate should finish that first - the same foundation serves both.
That is the work we do - see Fabric data platform work, and for the background, what Microsoft announced around it at Build 2026.
If you would like a straight assessment of where your estate stands, book a free 30-minute consultation.
Sources and Further Reading
Frequently asked
Rayfin is Microsoft's open-source SDK and command-line tool, announced at Build 2026, that lets developers and AI coding agents describe an application and get an enterprise-grade backend generated into the code - database, authentication and supporting services - deployed to Microsoft Fabric with the data in OneLake.
No. Asking questions of governed data in plain English is the job of Fabric IQ and Fabric Data Agents. Rayfin builds application backends: it is a development tool, not a BI or intelligence feature.
Because that puts the application's operational data inside the governed estate from day one. Tables land in OneLake next to your analytical data, Purview governance applies, and Power BI, Fabric IQ and Data Agents can use the application's data without a separate integration project.
Application developers and teams using AI coding agents who want to go from prototype to production without managing backend infrastructure - and, indirectly, the data teams who inherit the resulting data in OneLake.
It is new and open source, and generated backends deserve the same engineering review as hand-built ones: data modelling, security, and an operating model. Treat early output as a strong scaffold, and harden it the way you would any backend.
Operational data from Rayfin-built applications arrives in OneLake by default, which removes integration work but widens the governance surface. Capacity planning, workspace structure and access control need to account for application workloads alongside analytics.
