Faber Overview

Faber: Universal AI Agent Orchestration

Write-once, deploy-everywhere AI agent definitions.

Faber enables you to define AI agents once and deploy them to any framework (Claude Code, LangGraph, CrewAI) and any platform (GitHub, Linear, Jira, AWS, GCP) without rewriting code.

Key Features

🎯 Three-Dimensional Abstraction

1. Framework Abstraction - Deploy to Claude Code, LangGraph, CrewAI, or custom frameworks
2. Platform Abstraction - Work with GitHub, GitLab, Linear, Jira, AWS, GCP via MCP
3. Organization Abstraction - Customize for your company without forking

⚒️ Five First-Class Concepts

1. Roles - AI agent definitions with contexts, tasks, and flows
2. Tools - MCP servers and utilities for platform integration
3. Evals - Testing and evaluation scenarios
4. Teams - Multi-agent compositions
5. Workflows - Cross-team orchestrations

📚 Seven Context Categories

• Specialists - Domain expertise loaded on-demand
• Platforms - Platform-specific adapters (MCP)
• Standards - Best practices and conventions
• Patterns - Design patterns and architectures
• Playbooks - Operational procedures
• References - API and framework documentation
• Troubleshooting - Issue resolution guides

How It Works

1. Define agents in platform-agnostic YAML
2. Customize via overlays without forking
3. Build for any target framework
4. Deploy with framework-specific artifacts

Quick Example

roles/issue-manager/agent.yml

org: acme
system: devops
name: issue-manager
type: role
platforms: [github-issues, linear, jira]

# Build for different frameworks
fractary faber build claude role issue-manager     # → .claude/agents/
fractary faber build langgraph role issue-manager  # → langgraph/graphs/
fractary faber build crewai role issue-manager     # → crewai/agents/

Get Started

Installation → Quick Start Guide → Core Concepts →