Microsoft unveils Rayfin to turn Fabric into an enterprise AI application runtime

3 Sources

Share

Microsoft introduced Rayfin at Build 2026, an open-source SDK that lets developers and AI agents build application backends and deploy them directly to Microsoft Fabric. The tool addresses a critical gap in enterprise AI: preventing data silos while maintaining governance as coding agents generate applications faster than teams can manage them.

Microsoft Fabric Gets a Code-First Workflow for AI Applications

Microsoft unveiled Rayfin at Microsoft Build 2026, signaling a strategic push to transform Microsoft Fabric into a comprehensive AI application runtime

1

. The open-source SDK and CLI addresses what Shireesh Thota, CVP of databases at Microsoft, describes as the real challenge in AI-assisted development: operationalizing code rather than writing it. Rayfin enables developers and coding agents to define complete application backends in code, including databases, business logic, APIs, identity, and access policies, then deploy them as fully managed, enterprise-grade backends on Microsoft Fabric

1

.

Source: SiliconANGLE

Source: SiliconANGLE

The announcement comes as enterprises grapple with a critical problem: every AI agent deployed starts without memory of how the business operates, where data lives, or what rules apply. According to VentureBeat's VB Pulse Q1 2026 RAG Infrastructure Market Tracker, hybrid retrieval intent among organizations with over 100 employees tripled from 10.3% in January to 33.3% in March, indicating that enterprises have moved beyond expanding RAG coverage to focus on the underlying architecture

2

.

Eliminating Data Silos Through Unified Deployment

Rayfin tackles a widening gap in enterprise AI: as coding agents spin up applications at unprecedented speed, each one risks becoming another isolated silo outside the organization's data layer entirely. The tool deploys agent-built applications directly to Fabric as a governed production backend, routing application data into the same platform rather than creating new silos

2

. Once deployed, applications run as first-class artifacts inside Fabric and connect directly to OneLake, Microsoft's unified data layer, ensuring data governance and compliance come from Fabric itself rather than requiring per-app configuration

3

.

Microsoft positions Rayfin against Postgres-compatible backends like Supabase and Neon that agentic coding tools typically default to. The key differentiator is governance: Rayfin routes the entire application fleet through Fabric's unified data and compliance layer rather than creating isolated data silos

2

. Amir Netz, CTO of Microsoft Fabric, explained that the platform's role is "creating reality for agents based on data," emphasizing that the relationship is bidirectional—agents building Rayfin applications draw from the organization's ontology, and the data those applications generate enriches that ontology for subsequent agents

2

.

Microsoft IQ Provides Unified Context for Enterprise AI Agents

Source: VentureBeat

Source: VentureBeat

Alongside Rayfin, Microsoft expanded Fabric IQ into a broader system called Microsoft IQ, unifying four context sources into a single agent foundation. Work IQ captures organizational operations through email, documents, and meetings. Foundry IQ manages institutional knowledge and procedures. Fabric IQ models the live operational state through data, defining entities and relationships grounded in real-time signals. Web IQ adds real-time global context from the web

2

. This unified approach means developers can connect new agents to all four context sources in a single integration step, addressing the challenge that shared business context is what retrieval alone cannot solve.

Strongly Typed Definitions Enable Agent Collaboration

With Rayfin, developers or coding agents working on their behalf define data models, application programming interfaces, business logic, access policies, and connections to existing data sources in one place. The definitions are strongly typed, enabling coding agents like GitHub Copilot to read and modify them with the same reliability as a human developer

3

. The CLI handles deployment to Fabric and automatically provisions databases, authentication, access policies, and APIs—work that teams would otherwise wire together manually. Because applications sit inside the data platform, their data becomes immediately available for analytics and AI use the moment it's written, with no copies or pipelines required

3

.

Source: InfoWorld

Source: InfoWorld

Early Adoption and Platform Competition

Microsoft is pairing Rayfin with Replit, the AI-first coding platform, allowing developers to build in familiar environments while deploying into managed Fabric tenants. Amjad Masad, CEO of Replit, noted that the partnership gives developers "a path from idea to enterprise-grade production that's measured in hours, not months"

3

. Early customer Leatherman Tool Group is using the combination to build and iterate in Replit while keeping operational and analytical data centralized in Fabric

3

.

Microsoft isn't alone in building shared context layers for enterprise AI. Snowflake announced semantic capabilities this week, Pinecone expanded its vector database into the Nexus knowledge engine, and Redis developed its Iris context and memory platform. Robert Kramer, managing partner at KramerERP, observed that "the enterprise AI challenge is no longer just about model availability," questioning whether Microsoft will simplify execution and strengthen trust or add complexity to an already dense environment

2

. As deploying application backends becomes faster through AI-assisted development, the focus shifts to whether platforms can deliver unified data governance at the speed agents now operate.🟡 familiarity environment while deploying into managed Fabric tenants. Amjad Masad, CEO of Replit, noted that the partnership gives developers "a path from idea to enterprise-grade production that's measured in hours, not months"

3

. Early customer Leatherman Tool Group is using the combination to build and iterate in Replit while keeping operational and analytical data centralized in Fabric

3

.

Microsoft isn't alone in building shared context layers for enterprise AI. Snowflake announced semantic capabilities this week, Pinecone expanded its vector database into the Nexus knowledge engine, and Redis developed its Iris context and memory platform. Robert Kramer, managing partner at KramerERP, observed that "the enterprise AI challenge is no longer just about model availability," questioning whether Microsoft will simplify execution and strengthen trust or add complexity to an already dense environment

2

. As deploying application backends becomes faster through AI-assisted development, the focus shifts to whether platforms can deliver unified data governance at the speed agents now operate.

Today's Top Stories