What is Fabric Apps?

28th of July I wrote a LinkedIn post about Fabric Apps and thanks to Arun Ulag who reposted it with a comment “100%” it went viral. I think I received like 100+ new connections just from that single post. Anyway in that post I promised to write blog post about Fabric Apps and this is the first one of them.

Why I Love Fabric Apps?

Love might be strong word to use about technology, but I really-really like what Fabric Apps is bringing to the table. Fabric Apps is an answer to AI generated internal apps. It is a lot more than that, but if we think about the wide adoption case, I think that internal apps is the most common scenario for Fabric Apps (at least now).

Shadow AI

Shadow AI is the unsanctioned use of artificial intelligence tools, models, or embedded features by employees without the knowledge, approval, or oversight of an organization’s IT and security teams. It typically occurs via public chatbots, browser extensions, and personal accounts used to increase productivity.

Big companies struggles to introduce AI services to employees. Yes there are tools like Microsoft Copilot and Chat GPT Enterprise that are designed to do that, but these tools don’t solve the app infrastructure problem. More and more business users and “normal” workers want to build their own apps, that will improve their productivity on certain specific problems. App can be an hour checking tool for manager, currency value dashboard for market specialist or an ERP query builder for plant operator. Little tool that helps usually with one specific task. Generative AI tools like Claude can easily build these apps from simple prompts, but there are two problems that these tools don’t solve:

1. How can you share your application, so that your colleague can also run it?
2. How your app can SAFELY access the data from company systems like ERP?

These unsolvable problems can lead into use of shadow AI or shadow OPS. Well Fabric Apps solves both of these problems and in a very simple and secure way.

Fabric Apps

Fabric Apps is a new preview feature in Microsoft Fabric that allows users to create and run their own applications. These applications can easily access the company data from Fabric and they even support row level security. Row level security meaning that you can create apps where users don’t see other user data (as in my Todo App I only see my tasks).

Fabric Apps are created with a tool called Rayfin. Rayfin is a command line tool, but if you are afraid to use command line, you can also use free tool called Fabricator. It basically abstracts the Rayfin. In this post I won’t talk about Fabricator as it is big enough to have its own post.

The clever thing about Rayfin is, that you can create application without actually manually running it. Rayfin is a command line tool and AI tools are very good at using those, so you can simply put a prompt into Copilot or Claude and let the AI tool do the work.

Note: to create and run Fabric Apps you need to have Fabric capacity in a region where this feature is supported. For example West Europe is one of the supported regions.

Creating New Application

The sample prompt below creates a new Fabric App. You can type in your application specifications between bullets 5 and 6. Then just run the prompt with your AI tool and wait.

Set up a new Fabric App project called "My App".
1. Run: npm create @microsoft/rayfin@latest
2. Open the scaffolded project directory
3. Explore the project structure and understand the Rayfin app framework
4. Run npm run dev
5. Make changes to the app based on my instructions
6. When ready, deploy with: npx rayfin up

This is picture of my Todo sample application. This app was created with just a few prompts. Think about that this app is hosted in Fabric, it can access any data that I have stored into Fabric (and I have permission to access) and it supports writebacks. No manual setup at all.

So you can create, edit and delete data with Fabric Apps. All of this in a secure (Entra Authentication) and monitored environment.

Modifying the App

One thing that Fabric Apps does not solve yet is the version control. You need to store the source code somewhere and Fabric does not offer place for that. It offers place to run it, but it stores only the “compiled app”.

Most commonly you would use your company Azure DevOps or GitHub repository to store the code, but they require at least some understanding of a GIT version control system. I hope that we will see some improvements on this in near future, so that we could store the application source also inside Fabric.

If you want to modify your application you can simply do it with LLM prompts. Secret behind this is that every Fabric app project contains AGENTS.md and SKILL.md files, which are guidelines for AI tools to work with these apps and with Rayfin. By using these guidelines the AI knows how to update your app in Fabric. It knows how to add features and run the app in your own computer for testing purposes. You can use the Rayfin command line tool directly, but you don’t have to and that is the magic in Fabric Apps. It works completely through LLM tools.

Summary

OK, I don’t want to make the first part of this blog post series too long, so I will quit it here. If I had to summarize the current state of Fabric Apps in a single line it would be “Fabric Apps is the best way to build internal apps that safely use your organization’s data“. Is it something else in near future? Time will tell. I have some predictions about where all of this could go, but I don’t want to spill them out yet. However I can now say that there is a great promise in this tool and it is something that I eagerly wait to GA.