Agent Squad
A lightweight, open-source framework for orchestrating multiple specialized AI agents and routing conversations between them, in Python, TypeScript and now on-device Swift.
Agent Squad is a flexible, lightweight framework for orchestrating multiple AI agents, routing each user query to the most suitable specialized agent and maintaining conversation context across them. It provides pre-built agents, classifiers and storage for quick deployment, plus defined seams for plugging in custom ones.
A Classifier analyzes user input using the agents' descriptions and conversation history to select the best agent for a turn, the selected Agent processes the input and calls tools as needed, and the Orchestrator saves the exchange and returns the response. Pre-built agents cover Bedrock, Anthropic, OpenAI, Lex and Lambda, among others, with streaming and non-streaming responses.
Python and TypeScript maintain feature parity and run anywhere from AWS Lambda to a laptop, while a newer Swift runtime brings the same orchestration model to iOS and macOS, running entirely on device with local tools, native and MCP tool support, realtime voice, tracing and local-first chat storage. The project was formerly named multi-agent-orchestrator and hosted under a different organization.