• Getting Started
  • Core Concepts
  • Reinforcement Learning
  • Model Context Protocol (MCP)
  • Workflow Patterns
  • Advanced Agent Patterns
  • Guides

Getting Started

Overview

Overview of the az-core.

Arc is a professional agentic framework designed for building complex AI agent systems with advanced orchestration, reinforcement learning, and multi-agent reasoning — all with minimal code.

It enables developers to create hierarchical, intelligent agent workflows using a coordinator–planner–supervisor–team architecture, providing clear modularity and scalability for production systems.

With under 20 lines of code, you can design and run sophisticated multi-agent workflows capable of reasoning, coordination, and adaptive tool use.


🧩 Core Highlights

  • 🧠 Hierarchical Architecture — Built-in coordinator–planner–supervisor design for intelligent routing and multi-level agent collaboration.
  • ⚙️ Multiple Workflow Patterns — Includes sequential, concurrent, swarm, and hierarchical orchestration patterns optimized for various tasks.
  • 🎯 Reinforcement Learning Integration — Agents learn optimal tool usage over time using Q-learning and semantic state matching.
  • 🧠 Advanced Reasoning Patterns — Supports self-consistency, reflexion, reasoning duo, and evaluation-based agent systems.
  • 🌐 MCP Integration — Native Model Context Protocol support for connecting to external servers and real-world tools.
  • 🛡️ Production Ready — Includes caching, error handling, validation, visualization, monitoring, and retry logic.

🚀 When to Use Arc

Use Arc when building:

  • Multi-agent systems requiring hierarchical coordination
  • Complex workflows involving specialized agents and tools
  • Applications needing adaptive reasoning and continuous learning
  • Enterprise-grade or research-scale agent orchestration

💡 Example Use Cases

  • Research assistants combining multiple expert agents
  • Adaptive tool orchestration with reinforcement learning
  • Automated reasoning, planning, and evaluation pipelines
  • Scalable agent teams for enterprise or academic systems

Edit this page on GitHub
AzrienLabs logo

AzrienLabs

Craftedby Team AzrienLabs