Agent Squad is an open-source framework for managing multiple AI agents and the complex conversations they share. Originally created in the AWS ecosystem as the Multi-Agent Orchestrator, it provides the routing, memory and coordination layer needed when one chatbot is really a team of specialized agents behind the scenes. The project has about 7,800 stars on GitHub, is released under the Apache 2.0 license, and was actively developed in 2026.
The framework’s core idea is intelligent request classification: every user message is analyzed and routed to the agent best suited to answer it — a support agent, a sales agent, a code assistant — while conversation history stays coherent across switches. Agent Squad ships with built-in agents for Amazon Bedrock, Anthropic Claude, OpenAI and Lex, plus supervisor patterns that let one agent coordinate several others. It is available for both Python and TypeScript, runs comfortably on serverless infrastructure such as AWS Lambda, and gives you pluggable storage for conversation state. That makes it a pragmatic foundation for production multi-agent chat systems rather than a research demo.
Free. Agent Squad is open source under the Apache 2.0 license; you pay only for the model APIs and cloud infrastructure your agents use.
Developers building customer-facing assistants and internal copilots where different questions need different specialist agents, and who want routing, memory and orchestration handled by a proven framework.