CSA Loom — Solution Store Accelerator

CSA Loom is a productized Microsoft Fabric parity layer that deploys into any Azure tenant where Fabric isn't yet generally available. Federal, DoD, intelligence community, state + local, defense industrial base, and regulated commercial verticals.

What you get

• Loom Console — Next.js + Fluent UI v9 SaaS UI that mirrors the Microsoft Fabric workspace experience
• Loom Setup Wizard — conversational deploy with live .bicepparam preview backed by self-hosted Azure MCP
• Parity services that fill Fabric-only gaps (Direct-Lake Shim, Activator Engine, Mirroring Engine, Data Agents)
• Two-tier deployment — azd up CLI or "Deploy to Azure" button
• Forward-migration tooling — to Microsoft Fabric when your boundary reaches GA

Quickstart

60-minute Quick Start →

git clone https://github.com/fgarofalo56/csa-inabox.git
cd csa-inabox/platform/fiab/azd
azd init -t .
azd up

Architecture

Reference Architecture →

Per-boundary support

Boundary	v1 (now)	v1.1 (+3 mo)
Azure Commercial	✅	—
GCC	✅	—
GCC-High / IL4	✅	—
DoD IL5	—	✅

Cost

CSA Loom IP is free in v1. You pay only for Azure consumption underneath: - Sample F8 Commercial deployment: ~\(3-5K/month - Sample F8 GCC-High deployment: ~\)4-6K/month

Detailed cost breakdown →

Marketplace listing + pricing model deferred to backlog. See docs/fiab/deployment/marketplace.md for the future pricing roadmap placeholder.

Resources

• Documentation pillar
• Source code on GitHub
• Epic / build roadmap
• 5-day Cloud CoE workshops
• Marketing kit

How CSA Loom relates to other CSA accelerators

The API-First Data Strategy accelerator (see solution-store index) covers the integration layer (APIM + Dataverse + cross-platform APIs). CSA Loom covers the analytics + lakehouse + BI layer for audit-boundary-blocked customers.

Both pillars share the csa-inabox foundation (Bicep, Copilot backend, MkDocs Material).

Forward migration

CSA Loom forward-migrates 1:1 to Microsoft Fabric when Fabric reaches your audit boundary. Delta tables via OneLake shortcut (zero data movement); dbt + KQL port unchanged; semantic models re-author for Direct Lake on OneLake. See forward-migration runbook.