Agent Squad

Agent Squad is a flexible, open-source multi-agent orchestration framework available in both Python and TypeScript, maintained by 2FastLabs under the Apache License 2.0. Originally hosted at awslabs/agent-squad, the project has moved to 2fastlabs/agent-squad and continues active development. The framework intelligently routes user queries to the most suitable agent based on context and conversation history, making it suitable for everything from simple chatbots to sophisticated enterprise AI systems.

What It Is

Agent Squad is a multi-agent orchestration layer that sits between user input and a collection of specialized AI agents. A built-in Classifier analyzes each incoming query, consults agent characteristics and conversation history, selects the best-fit agent, processes the response, and saves the updated conversation state — all before returning the answer to the user. The framework supports both streaming and non-streaming responses and is designed to run anywhere, from AWS Lambda to local environments or any cloud platform.

Core Architecture

The system follows a clear flow: user input → Classifier → agent selection → agent processing → conversation storage → response delivery. Key architectural elements include:

• Intelligent intent classification — dynamically routes queries to the most suitable agent based on context and content
• Context management — maintains conversation history across multiple agents for coherent multi-turn interactions
• SupervisorAgent — a newer "agent-as-tools" architecture that enables a lead agent to coordinate a team of specialized sub-agents in parallel, supporting dynamic delegation and smart context management
• Extensible architecture — custom agents and conversation storage solutions can be integrated alongside pre-built components

Built-In Agents and Integrations

The framework ships with a variety of ready-to-use agents and classifier implementations:

• BedrockLLMAgent — Amazon Bedrock LLM with Converse API support
• AmazonBedrockAgent — full Amazon Bedrock Agent integration
• BedrockInlineAgent — uses the InvokeInlineAgent API for dynamic agent creation
• BedrockFlowsAgent — integrates Amazon Bedrock Flows with persistent memory
• OpenAI Agent — GPT model support for agents and classification
• LexBotAgent — Amazon Lex bot integration for specialized dialog tasks
• LambdaAgent — AWS Lambda function as an agent
• BedrockTranslatorAgent, ComprehendFilterAgent, ChainAgent — additional specialized built-ins

Modular Python installation options ([aws], [anthropic], [openai], [all]) let developers include only the dependencies they need.

Deployment and Setup

Agent Squad is installable via npm (agent-squad) for TypeScript and via pip (agent-squad[aws] or other extras) for Python. The documentation covers local development, AWS Lambda deployment for both languages, Chainlit chat app integration, and FastAPI streaming setups. Example projects in the repository include a web-based chat demo, an e-commerce support simulator with human-in-the-loop routing, a Streamlit global demo featuring an AI Movie Production Studio and AI Travel Planner, and a FastAPI streaming implementation.

Update: python_1.0.2

The latest published release is python_1.0.2, published on 2025-06-25, reflecting continued active maintenance after the project's migration from awslabs/agent-squad to 2fastlabs/agent-squad. The repository was created in July 2024 and last pushed in May 2026, with the GitHub repository reporting over 7,600 stars and 719 forks according to the project's own badges. The project follows an issue-first pull request policy and uses GitHub Actions to enforce PR-to-issue linking.