Agent Squad

Agent Squad (formerly Multi-Agent Orchestrator) by AWS Labs is a flexible, lightweight, open-source framework for managing and orchestrating multiple AI agents to handle complex, multi-turn conversations and tasks intelligently. It uses a powerful classification system that dynamically routes user queries to the most suitable specialist agent based on context, conversation history, and agent capabilities. Designed to support diverse AI architectures and deployments, Agent Squad enables seamless collaboration between specialized agents, maintaining conversation context across interactions. It serves a wide range of applications from simple chatbots to sophisticated multi-agent AI systems, scaling efficiently and adaptable to various cloud or local environments.

Key Features:

• Intelligent Intent Classification: Dynamically routes queries to the appropriate agent by analyzing conversation context, user input, and agent descriptions.
• Context Management Across Agents: Maintains and uses conversation history from multiple agents to ensure coherent and continuous interactions.
• Extensible and Modular Architecture: Easily integrate or customize agents and classifiers, supporting a wide variety of
• AI models and tools (e.g., Amazon Bedrock, Lex bots).
• Universal Deployment: Runs anywhere including AWS Lambda, local environments, containers, and edge computing platforms.

Use Cases:

• Customer support systems with specialized agents for different inquiry types (e.g., travel, health, tech support).
• Complex conversational AI applications such as travel booking involving multiple services, contextual weather updates, and real-time recommendations.
• Multi-agent coordination for product development, healthcare triage, and e-commerce support with intelligent query routing and workflow automation.

Technical Specifications:

• Dual implementation in Python and TypeScript, supporting flexible integration depending on development needs.
• Built-in classifiers using large language models to analyze queries and maintain conversation state with scalable storage options (in-memory, DynamoDB).
• Supports streaming and non-streaming responses, parallel processing, and can coordinate multiple agents simultaneously using advanced orchestration components like SupervisorAgent.