Agent Squad is a flexible, lightweight framework for orchestrating multiple AI agents to handle complex conversations. It intelligently routes queries, maintains context across interactions, and supports both Python and TypeScript. This framework is ideal for developers building sophisticated AI systems, chatbots, or customer support solutions requiring multi-agent coordination.

How It Works

The core of Agent Squad is its orchestrator, which uses a classifier to intelligently route user queries to the most suitable agent. This classification leverages agent characteristics and conversation history. A key feature is the SupervisorAgent, enabling "agent-as-tools" coordination where a lead agent can delegate subtasks to specialized agents in parallel, maintaining context for coherent responses. This architecture supports dynamic delegation and compatibility with various agent types (e.g., Bedrock, Anthropic, Lex).

Quick Start & Requirements

• Python: pip install agent-squad[aws] (or [anthropic], [openai], [all])
• TypeScript: npm install agent-squad
• Prerequisites: AWS credentials (if using [aws]), specific LLM API keys depending on agent configuration.
• Documentation: Full Documentation

Highlighted Details

• Dual language support: Python and TypeScript implementations available.
• SupervisorAgent for sophisticated team coordination and parallel processing.
• Flexible agent responses: supports streaming and non-streaming outputs.
• Universal deployment: runs on AWS Lambda, local environments, or any cloud platform.

Maintenance & Community

The project is actively maintained by AWS Labs. Community engagement is encouraged through GitHub Issues for discussions, ideas, and pull requests.

Licensing & Compatibility

Licensed under the Apache 2.0 license. This license is permissive and generally compatible with commercial and closed-source applications.

Limitations & Caveats

The project was recently renamed from "Multi-Agent Orchestrator" to "Agent Squad," which might lead to some older documentation or discussions using the previous name. Contributions require linking to an existing GitHub issue.