How Fabric Works
A product-level overview of how Fabric AI processes requests, generates documents, and orchestrates agents.
Fabric AI combines intelligent agents, durable workflows, and document retrieval to help teams automate complex knowledge work. This page explains the key concepts behind how the platform works.
Request Flow
When you interact with Fabric, your request flows through several stages:
- You submit a request through chat, a workflow trigger, or the API
- The Orchestrator determines which specialized agent is best suited for your task
- The agent retrieves context from your uploaded documents and workspace knowledge
- AI generates the result using your configured provider (OpenAI, Anthropic, Groq, etc.)
- You receive the output as a document, chat response, or workflow result
AI Agents
Agents are the intelligent core of Fabric. Each agent specializes in a type of task:
| Agent | What It Does |
|---|---|
| Orchestrator | Routes your requests to the right specialist and coordinates multi-agent tasks |
| Document Generator | Creates PRDs, specs, architecture docs, and other structured documents |
| CUGA | Configurable generalist agent for complex, multi-step automation |
| Custom Agents | Agents you define with specific tools, prompts, and behaviors |
Agents can communicate with each other, use external tools (via MCP), and learn from past interactions through persistent memory.
Document Retrieval (RAG)
When you upload documents to a workspace, Fabric processes them into searchable knowledge:
- Documents are chunked into meaningful segments
- Embeddings are generated for semantic understanding
- When you ask a question, relevant chunks are retrieved and provided as context to the AI
This means the AI references your actual documents rather than relying solely on its training data. See RAG for more details.
Durable Workflows
Fabric uses a durable workflow engine to handle long-running, multi-step processes:
- Fault tolerance -- If a step fails, it retries automatically
- Human-in-the-loop -- Workflows can pause for your approval before continuing
- State persistence -- Workflows survive restarts and interruptions
- Audit trail -- Every step is recorded for full visibility
This powers features like document generation pipelines, scheduled agent runs, and multi-tool automation chains. See Workflows for more details.
Built-In Integrations
Fabric connects to your existing tools to automate end-to-end processes:
| Integration | What You Can Do |
|---|---|
| Slack | Send messages and notifications |
| GitHub | Create issues and pull requests |
| Linear | Create and update issues |
| Send automated emails | |
| Webhooks | Connect to any external system |
Additional integrations are available through MCP (Model Context Protocol). See Integrations for the full list.
Content Processing Tools
Fabric includes built-in tools for processing content from various sources:
- YouTube -- Extract transcripts and analyze video content
- Web -- Scrape and analyze web page content
- Audio -- Transcribe audio files
Data Isolation
All data in Fabric is strictly isolated by tenant. Your personal data is never visible in an organization, and one organization's data is never visible to another -- even if the same user belongs to both. See Data Isolation for details.
Security
- Multiple sign-in methods -- Email, OAuth, passkeys, and two-factor authentication
- Role-based access -- Owner, Admin, and Member roles within organizations
- Encrypted credentials -- All API keys and secrets are encrypted at rest
- Encrypted communications -- All data in transit is protected with TLS