What is a Fabric? The Complete Guide to AI-Powered Tools
If you've ever wished AI could just do the thing instead of requiring endless prompting and tweaking, you're going to love Fabrics.
A Fabric is a self-contained, purpose-built AI tool that does one thing really well. Unlike general-purpose chatbots where you have to explain what you want every single time, a Fabric already knows its job. You provide the inputs, and it delivers the outputs. Simple.
Think of it this way: ChatGPT is like having a brilliant assistant who can do anything—but you have to explain the task from scratch every time. A Fabric is like having a specialist tool that's been pre-configured for a specific job. You wouldn't use a Swiss Army knife to cut a steak when you have a steak knife right there.
In this guide, we'll break down everything you need to know about Fabrics: what they are, how they work, and why they're changing how people interact with AI.
Table of Contents
- The Problem Fabrics Solve
- Anatomy of a Fabric
- The 11 Fabric Types
- How Fabrics Work
- The Refinement System
- Fabrics vs Traditional AI Tools
- Getting Started
The Problem Fabrics Solve
Let's be honest: working with AI can be frustrating.
You open ChatGPT, write a detailed prompt explaining exactly what you want, wait for the response, realize it's not quite right, rewrite the prompt, wait again, tweak the output manually, and repeat. For a simple task like "write me a product description," you might spend 15 minutes getting something usable.
Now multiply that by every task you need to do: social media posts, email drafts, image generation, code snippets, data analysis. The "time-saving" AI tool suddenly consumes hours of your day just in prompt engineering.
Fabrics solve this by front-loading the complexity.
When someone creates a Fabric, they've already figured out:
- What inputs are needed
- How to structure the prompt
- Which AI model works best for this task
- How to format the output
- What refinement options make sense
All you have to do is fill in the blanks and click "Run."
A well-designed Fabric turns a 15-minute prompting session into a 30-second form submission.
Anatomy of a Fabric
Every Fabric is built from the same core components. Understanding these will help you use Fabrics more effectively—and eventually create your own.
1. Metadata
The Fabric's identity: its name, description, category, and tags. This is what helps you find the right Fabric for your task when browsing the marketplace.
2. Inputs
The data you provide. Inputs are typed fields—text, numbers, images, files, selections—organized into a form. Each input has:
- Type: What kind of data (text, number, image, etc.)
- Label: What the field is for
- Validation: Rules the input must follow
- Default values: Pre-filled suggestions
- Conditional visibility: Some fields only appear based on other selections
For example, a "Blog Post Generator" Fabric might have inputs for:
- Topic (text)
- Tone (select: professional, casual, witty)
- Word count (number slider)
- Target audience (text)
- Include images (toggle)
The input system supports over 20 field types, from simple text boxes to complex structured lists and code editors. Learn more in our complete guide to Fabric inputs.
3. Logic
The system prompt and instructions that tell the AI how to process your inputs. This is the "secret sauce" that makes each Fabric unique.
The logic layer is model-agnostic, meaning it can work with different AI models (Claude, GPT-4, Gemini, etc.) without modification. The Fabric creator has optimized this prompt so you don't have to.
4. Capabilities
What the Fabric can actually do. Each Fabric has 1-4 capabilities from this list:
- Generate: Create new content from scratch
- Transform: Rewrite or convert existing content
- Extract: Pull insights or data from content
- Classify: Categorize or label content
- Summarize: Condense long content
- Structure: Organize into a specific format
- Media Process: Work with images, video, or audio
- Plan: Create workflows or strategies
- Validate: Check content against rules
5. Outputs
What you get back. Fabrics support over 40 output types:
Text & Documents
- Plain text, Markdown, structured documents
- Essays, guides, blueprints
Data & Structured
- JSON, tables, analysis reports
- Charts and diagrams
Media
- Images (single, gallery, carousel)
- Videos, audio files
- 3D models, animations
Code
- Code files in any language
- React components, HTML, SVG
The output is presented in a display optimized for its type—images show in a gallery, code gets syntax highlighting, tables are properly formatted.
6. Configuration
Settings that control how the Fabric runs:
- Model selection: Which AI model to use
- Temperature: How creative vs. deterministic
- Max tokens: Output length limits
- Refinement settings: What can be adjusted after generation
The 11 Fabric Types
Not all Fabrics are created equal. We categorize them into 11 types based on their primary function:
| Type | What It Does | Example Use Case |
|---|---|---|
| Generate | Creates new content | Blog post writer, logo generator |
| Transform | Rewrites or converts | Tone converter, format translator |
| Analyze | Extracts insights | Sentiment analyzer, data extractor |
| Build | Produces code/assets | Component builder, API generator |
| Research | Synthesizes information | Report compiler, literature reviewer |
| Plan | Creates strategies | Content calendar, project planner |
| Utilities | Validates/calculates | Grammar checker, unit converter |
| Enhance | Improves quality | Image upscaler, audio enhancer |
| Clone | Replicates styles | Voice cloner, style transfer |
| Segment | Detects/masks | Background remover, object detector |
| Compose | Combines elements | Avatar creator, lipsync generator |
Each type has recommended output formats and capabilities that make sense for its purpose. A "Generate" Fabric naturally outputs text or media. An "Analyze" Fabric typically outputs JSON or structured reports.
Read the full breakdown in The 11 Fabric Types Explained.
How Fabrics Work
Here's what happens when you run a Fabric:
Step 1: Input Collection
You fill out the form with your requirements. The Fabric validates your inputs, ensuring everything is correct before proceeding.
Step 2: Prompt Construction
The Fabric takes your inputs and weaves them into its pre-built logic/prompt. This creates a complete, optimized request for the AI model.
Step 3: Model Routing
The Fabric determines which AI model to use. Some Fabrics are locked to a specific model; others intelligently route based on:
- The type of output needed
- The complexity of the task
- Your preferences
- Cost optimization
Step 4: Execution
The request goes to the AI model. For text outputs, this is typically Claude, GPT-4, or Gemini. For images, it might be Flux, Ideogram, or DALL-E. For video, Sora, Veo, or Kling.
Step 5: Output Processing
The raw AI response is parsed, formatted, and presented according to the Fabric's output configuration. Images display in a gallery, code gets syntax highlighting, data populates into tables.
Step 6: Refinement (Optional)
Not happy with the result? Use the refinement system to adjust the output without starting over.
The Refinement System
This is where Fabrics really shine over traditional AI tools.
The refinement system uses a two-phase approach:
Phase 1: Initial Generation
You provide your inputs, and the Fabric generates the first output. This is your baseline.
Phase 2: Iterative Refinement
Instead of regenerating from scratch, you can adjust the output using refinement controls specific to that output type.
For text outputs, you might adjust:
- Tone (more formal, more casual)
- Length (expand, condense)
- Complexity (simplify, add detail)
- Focus (emphasize different aspects)
For image outputs, you might adjust:
- Style (more realistic, more artistic)
- Composition (zoom, reframe)
- Colors (warmer, cooler)
- Details (add, remove elements)
The Lock System
Here's the magic: you can lock parts of the output you like.
Say the Fabric generated a product description, and the first paragraph is perfect but the second needs work. Lock the first paragraph, then refine only the second. The AI will preserve exactly what you locked while improving the rest.
Lock groups let you protect related elements together:
- Lock "Content" to keep the message but change the style
- Lock "Tone" to keep the voice but change the content
- Lock "Structure" to keep the format but change the details
Refinement Layers
Every refinement creates a new "layer"—a snapshot of that version. You can:
- Undo to previous versions
- Compare different iterations
- Branch off in different directions
It's like version control for AI outputs.
Quick Actions
Fabric creators can define one-click refinement buttons:
- "Make it shorter"
- "More professional"
- "Add statistics"
- "Different angle"
These inject pre-configured refinement prompts, saving you from typing the same adjustments repeatedly.
Learn more in The Refinement System: Edit AI Outputs.
Fabrics vs Traditional AI Tools
| Aspect | Traditional AI (ChatGPT, etc.) | Fabrics |
|---|---|---|
| Setup time | Write prompt from scratch | Fill out a form |
| Consistency | Varies by prompt quality | Consistent results |
| Learning curve | High (prompt engineering) | Low (structured inputs) |
| Refinement | Regenerate or manual edit | Targeted adjustments with locks |
| Specialization | General purpose | Purpose-built for specific tasks |
| Sharing | Share prompts (fragile) | Share complete tools |
| Output format | Plain text | 40+ formatted types |
Fabrics aren't replacing general-purpose AI—they're complementing it. Use ChatGPT or Ask when you need flexibility. Use Fabrics when you have a defined task and want speed and consistency.
Getting Started
Ready to try Fabrics? Here's your path:
1. Browse the Marketplace
Explore hundreds of pre-built Fabrics across categories:
- Design (logos, graphics, mockups)
- Writing (blogs, emails, copy)
- Development (code, APIs, components)
- Marketing (ads, social posts, SEO)
- Video (scripts, thumbnails, editing)
- And many more
2. Run Your First Fabric
Pick a Fabric that matches a task you do regularly. Fill in the inputs, click "Run," and see the result. Try the refinement controls to adjust the output.
3. Save Your Preferences
Many Fabrics have "config" inputs—settings like your brand voice, preferred tone, or default style. These save with your preferences so you don't have to re-enter them every time.
4. Create Your Own
Once you understand how Fabrics work, create your own! Use the AI-assisted generator or the manual builder to design Fabrics for your specific workflows.
Learn more in How to Create a Fabric in 60 Seconds.
What's Next?
Now that you understand what Fabrics are, explore deeper:
- The 11 Fabric Types Explained - Learn which type fits your use case
- Fabric Inputs: Complete Guide - Master the input system
- Fabric Outputs: 40+ Types - Understand all output formats
- Fabrics for Designers - See role-specific examples
- Fabric Chains: Connect Workflows - Build multi-step automations
Fabrics are changing how people work with AI. Instead of fighting with prompts, you work with purpose-built tools. Instead of starting from scratch, you build on proven foundations. Instead of hoping for good outputs, you refine your way to perfect results.
Welcome to the future of AI workflows.