Rayfin on Microsoft Fabric: The End of Backend Boilerplate for AI Apps

On June 2, 2026 — during the opening keynote of Microsoft Build — a new name entered the developer lexicon: Rayfin. It didn’t ship as a service you click together in the Azure portal. It shipped as three terminal commands. And for developers building AI-powered applications, it changes the backend equation entirely.

This post breaks down what Rayfin is, how it works under the hood, what it means specifically for developers, and which new classes of applications it makes possible.

The Problem Rayfin Was Built to Solve

Vibe coding — the practice of describing what you want in natural language and having an AI agent generate it — has made the front-end almost trivially fast to produce. A polished UI can emerge in minutes. But the backend? That’s still a multi-week project of stitching together a database, an API layer, authentication, storage, governance, and deployment pipelines — and then wiring all of it into your existing data estate.

The gap shows up hard in enterprise settings. An agent can scaffold a beautiful React app, but the moment you ask it to “also set up the backend,” you’re back to writing Terraform, configuring managed identities, arguing with firewall rules, and manually onboarding into your data governance platform. The productivity promise of AI-first development collapses at the infrastructure boundary.

“Agents write the code. Fabric ships it quickly and safely.”

Amjad Masad, CEO of Replit — Microsoft Build 2026

Rayfin is Microsoft’s answer: a Backend-as-a-Service (BaaS) that collapses the entire infrastructure side of a Fabric-hosted application into code your team (or your agent) writes — and deploys in one command.

What Is Rayfin?

Rayfin is an open-source SDK and CLI — currently in public preview — that lets developers and AI coding agents define a complete application backend in code and deploy it as a first-class artifact on Microsoft Fabric.

What does “complete backend” mean here? When you run npx rayfin up, Fabric automatically provisions and manages:

• ✅ A SQL database — schema defined via TypeScript decorators, no manual DDL
• ✅ Microsoft Entra ID authentication — identity baked in from day one
• ✅ A schema-driven GraphQL API — auto-generated from your data model
• ✅ OneLake-connected storage — app data lands in OneLake automatically
• ✅ Static content hosting — serve your frontend directly from Fabric
• ✅ Governance and compliance — inherited from your Fabric tenant’s Purview and DLP policies

All of this from a single rayfin up command. No portal clicks. No Bicep templates. No manual role assignments.

How Rayfin Works: The Developer Workflow

Rayfin’s developer experience is deliberately minimal. The full lifecycle is:

Step 1 — Scaffold

npm create @microsoft/rayfin@latest

This generates a project with the TypeScript SDK, a rayfin.yml configuration file, and example data models. The entire backend is described in this scaffolded project — no separate infrastructure files.

Step 2 — Define Your Backend in Code

Data models are defined using TypeScript decorators. Rayfin reads these definitions to generate the database schema, GraphQL API, and access policies — all strongly typed and agent-friendly:

import { Entity, Field, PrimaryKey } from '@microsoft/rayfin-core'
@Entity()
export class Order {
  @PrimaryKey()
  id: string
  @Field()
  customerId: string
  @Field()
  amount: number
  @Field()
  status: 'pending' | 'shipped' | 'delivered'
  @Field()
  createdAt: Date
}

From this definition, Rayfin automatically generates your SQL table, a GraphQL endpoint to query and mutate orders, and wires it into OneLake — all before you run a single deploy command.

Step 3 — Deploy

npx rayfin up

This single command provisions everything in your Fabric workspace. Fabric validates the deployment contract, applies your Purview governance rules, and returns a live endpoint. The app is in production — governed, secured, and analytics-ready — without a single manual infrastructure step.

Local Development

Rayfin ships with an experimental local development mode that requires no cloud resources. You can iterate on your data models and business logic offline, then push to Fabric when ready. This makes it viable for laptop-first development workflows, even in air-gapped environments.

What This Means for Developers

Rayfin isn’t just a convenience layer — it’s a shift in what developers are responsible for. Here’s the concrete impact:

1. Infrastructure Is No Longer Your Problem

Database provisioning, connection strings, managed identities, network security groups, API gateway configuration — all of this disappears. Rayfin and Fabric own the infrastructure layer. You write application code.

2. Governance Ships With the App, Not After

In traditional workflows, compliance and governance get bolted on at the end — often painfully. With Rayfin, every app deployed to Fabric automatically inherits the tenant’s Microsoft Purview policies, DLP rules, and audit logging. There’s no “governance phase” because Fabric validates policy compliance as part of the deploy contract.

3. Your App Data Is Instantly Part of Your Data Estate

Because Rayfin stores application data in OneLake by default, the moment a record is written by your app, it is available — without ETL, without pipelines, without copies — to Power BI for reporting, to Fabric notebooks for data science, and to Fabric data agents for AI workflows. The line between operational data and analytical data disappears.

4. Coding Agents Can Operate the Full Stack

Rayfin’s TypeScript-first, strongly-typed design is explicitly agent-friendly. GitHub Copilot, Replit Agent, and other coding agents can read, extend, and refactor Rayfin backends with confidence — because the framework enforces structure. An agent that defines a new entity and runs npx rayfin up just deployed a production backend. That’s a fundamentally new capability.

5. Time-to-Production Drops from Months to Hours

Early adopters report going from “months” to “hours” for backend setup. Leatherman’s AI Solutions Architect noted that Rayfin enables fast development in preferred tools while keeping data “governed and centralized” on Fabric — something that previously required a dedicated platform engineering team to achieve.

What Rayfin Unlocks: New Classes of Applications

The combination of Rayfin + Microsoft Fabric enables application patterns that were previously either impossible or required significant platform engineering investment:

Transactional + Analytical Apps on One Platform

Traditionally, you ran a transactional database (OLTP) and a separate analytical warehouse (OLAP) and moved data between them with ETL pipelines. Rayfin collapses this by writing operational data directly to OneLake. You can build a customer-facing order management app and have that same data powering your executive Power BI dashboard — live, with no sync delay.

AI-Native Applications

Because Rayfin apps run inside Fabric, they have native access to Fabric’s AI capabilities — data agents, real-time intelligence, and ML models — without building integration plumbing. An app can write data, trigger a Fabric data agent workflow, and surface AI-generated insights back to the user, all within the same governed boundary.

Agent-Built Production Backends

The Replit partnership demonstrates the most radical use case: Replit Agent describes a user’s requirements, defines the Rayfin backend in code, runs npx rayfin up, and delivers a governed production endpoint — without a human developer touching the infrastructure. The agent is a full-stack developer.

Enterprise Apps Built by Non-Platform Teams

Rayfin removes the dependency on platform engineering teams to get a new app production-ready. A data science team, a business intelligence team, or even a business analyst with Replit access can now ship a governed internal application on Fabric — without needing an infrastructure specialist.

The Replit Partnership: Agents All the Way Down

Replit is the exclusive launch partner for Rayfin. The pairing is deliberate: Replit is where vibe coding happens at scale, and Rayfin is what makes those apps enterprise-safe. The flow looks like this:

1. User describes an app to Replit Agent in natural language
2. Replit Agent generates the front-end and defines the Rayfin backend in TypeScript
3. Replit Agent runs npx rayfin up — deploying to the user’s Fabric workspace
4. The app is live, governed, and analytics-ready in the enterprise’s Fabric tenant

The result is a complete path from idea to enterprise-grade production — measured in hours, not months — with no human infrastructure work. The developer (or agent) stays in the application layer. Fabric owns the rest.

Getting Started with Rayfin

Rayfin is currently in public preview, available to any organization with a Microsoft Fabric workspace. No credit card required — you can start with a free Fabric trial.

# 1. Create a new Rayfin project
npm create @microsoft/rayfin@latest my-app
# 2. Define your data models in TypeScript
cd my-app
# Edit src/models/ to define your entities
# 3. Deploy to your Fabric workspace
npx rayfin up

The SDK is open source at github.com/microsoft/rayfin with 14+ npm packages covering the CLI, core framework, client SDKs, authentication helpers, storage utilities, and Fabric integration. Documentation is available on Microsoft Learn.

The Bigger Picture

Rayfin is the clearest signal yet that Microsoft is positioning Fabric not just as a data analytics platform, but as a runtime for enterprise applications. By making Fabric the target for AI-generated backends, Microsoft is betting that the next generation of enterprise software gets written — and deployed — by agents, not by operations teams.

For developers, the message is straightforward: if you’re building anything that touches enterprise data, Rayfin removes the biggest friction point between idea and production. The infrastructure is managed. The governance is automatic. The analytics are live from day one.

The backend boilerplate era is ending. Rayfin just made the case for what comes next.

Rayfin is currently in public preview. Try it at microsoft.com/en-us/microsoft-fabric/features/rayfin or explore the open-source SDK at github.com/microsoft/rayfin.