Microsoft Rayfin & Fabric Apps: Complete Developer Guide

Why Microsoft Built Rayfin

AI coding agents have made it trivial to build a working application frontend in under an hour. The backend required to run that application in enterprise production is still a multi-week ordeal — auth configuration, database setup, access policies, compliance wiring, and governance controls that almost always get bolted on too late.

This is the gap Microsoft is closing with Fabric Apps and the Rayfin SDK. The bottleneck in 2026 is not model capability — it is consistent, shared data context. Every new agent or app starts from zero, relearning where data lives and what rules to follow. Without a consistent foundation, agents can’t coordinate and apps create silos.

What Is Microsoft Rayfin?

Rayfin is an open-source SDK and CLI that turns backend development into a code-first, declaration-driven workflow. Developers — or AI coding agents working on their behalf — describe the application backend in TypeScript. Rayfin generates the infrastructure and deploys it directly to Microsoft Fabric.

Once deployed, the application runs as a first-class Fabric artifact — visible in the Fabric portal, connected to OneLake, and governed by the same security and compliance controls already enforced across the tenant. Application data lands in OneLake automatically. No copy. No ETL pipeline. No synchronization delay.

The Replit partnership is the commercial headline: Replit is the exclusive launch partner. Developers build in Replit’s AI coding environment, deploy with Rayfin, and the entire application — code, data, and services — stays managed inside the organisation’s own Fabric tenant.

Fabric Apps Architecture

A Fabric App deployed to a workspace provisions three managed child services. These are not separate resources you configure — Fabric creates and manages them automatically based on your rayfin.yml configuration.

App Backend Endpoint Structure

Each Fabric App gets a single base endpoint. All services are exposed through path-based routing:

https://-app.rayfin.windows.net/

TypeScript Data Models & GraphQL

This is where Rayfin’s developer experience stands apart. You define your data model once in TypeScript using decorators. Rayfin generates the SQL schema, the GraphQL API, and the type-safe client — all automatically. No separate ORM configuration. No hand-written schema migrations.

// From Microsoft Learn official documentation
import {
  entity, role, text, boolean, date, uuid
} from '@microsoft/rayfin-core';

@entity()
@role('authenticated', '*', {
  policy: (claims, item) => claims.sub.eq(item.user_id),
})
export class Todo {
  @uuid()   id!: string;
  @text()  title!: string;
  @boolean() completed!: boolean;
  @date()  createdAt!: Date;
  @uuid()  user_id!: string;
}

From this TypeScript class, Rayfin automatically generates: a SQL table with the correct column types, a GraphQL endpoint with full CRUD operations, a row-level security policy based on the @role decorator, and a type-safe client SDK with autocomplete.

import { RayfinClient } from '@microsoft/rayfin-client';
import type { Todo } from '../rayfin/data/Todo';

const client = new RayfinClient<{ Todo: Todo }>({
  baseUrl: import.meta.env.VITE_RAYFIN_API_URL,
  publishableKey: 'pk-your-project-key',
});

// Type-safe query — autocomplete on all fields
const todos = await client.data.Todo
  .select(['id', 'title', 'completed', 'createdAt'])
  .execute();

// Find by primary key
const todo = await client.data.Todo
  .findByPk('00000000-0000-0000-0000-000000000000');

CLI Reference & Deployment Workflow

The Rayfin CLI handles the complete development lifecycle from scaffolding to production deployment. Install it once and the three core commands cover the majority of your workflow.

npm i @microsoft/rayfin-cli

• Scaffold a new project Creates a new project from a template — includes a React frontend, TypeScript data models, and rayfin.yml configuration. Choose from Blank App, To-Do App, or Data App templates.
• Run locally with Docker The full stack runs locally — frontend dev server, local database, and auth emulation. Rapid iteration before committing to a Fabric deployment.
• Deploy to Microsoft Fabric One command compiles frontend assets, evaluates schema changes, and provisions or updates all managed child services in the Fabric workspace.

# 1. Create a new project from template
npm create @microsoft/rayfin@latest my-app \
  --workspace 

# 2. Run frontend dev server locally
cd my-app
npm run dev

# 3. Deploy full application to Microsoft Fabric
npx rayfin up

# Apply TypeScript schema changes to Fabric SQL database
npx rayfin up db apply

# For existing projects (add Rayfin to existing codebase)
npx rayfin init my-app \
  --services db,storage \
  --auth-methods fabric \
  --static-hosting

# Check deployment status
npx rayfin status

Rayfin vs Supabase, Neon, Power Apps & Azure App Service

Microsoft explicitly positions Rayfin against Supabase and Neon — the Postgres-compatible backends that AI coding tools default to. Understanding the real trade-offs matters before committing your architecture.

Rayfin vs Supabase & Neon

Rayfin vs Power Apps

Real Enterprise Use Cases

Supply Chain & Field Operations

Field workers log delivery or inspection status through a React frontend. Writes go directly to the Fabric SQL database via GraphQL mutations. That data is immediately queryable by Power BI semantic models and AI agents without any ETL delay — on-time delivery metrics update in real time rather than with a multi-hour lag from a nightly pipeline.

Agentic AI Apps with Shared Business Context

AI coding agents building Rayfin applications draw from the organisation’s existing Fabric IQ context layer — the unified business ontology covering data, semantics, and real-time signals. (We break down this workflow in our Agentic Data Engineering Tutorial). App data generated by those agents feeds back into the same context layer, enriching it for subsequent agents. This bidirectional flow is what prevents the data silo problem that affects every other agentic backend approach.

Custom Executive & Operational Dashboards

When standard Power BI layout constraints prevent exact KPI presentations, Fabric Data Apps allow developers to build pixel-perfect React dashboards that execute DAX queries directly against trusted semantic models. (See the new PBIR enhanced report format capabilities for more on this shift). Visualization as Code — storing DAX and React TSX files in Git — gives these dashboards the same version control and deployment rigour as the rest of the application.

Licensing, Pricing & Capacity Requirements

Rayfin does not have its own pricing tier. Fabric Apps consume Capacity Units (CUs) from your organisation’s existing F-SKU capacity. This means the cost is additive to whatever you’re already spending on Fabric compute.

Frequently Asked Questions