agent-squad
A flexible and lightweight multi-agent orchestration framework that supports both cloud and on-device execution.
Install & Use
Copy this prompt and send it to your AI assistant (Claude / Cursor / TRAE / Codex / WorkBuddy etc.) to auto-install:
Help me install this AI Skill: agent-squad. It is used for: A flexible and lightweight multi-agent orchestration framework that supports both cloud and on-device execution. Full Skill content: https://321skill.com/skills/agent-squad/raw/index.md Read that page and install it.
The prompt includes a link to the full Skill content. You can also view the full content.
Agent Squad is an open-source framework designed to address the limitations of a single AI model in handling complex or cross-domain tasks. It employs an intelligent classifier to automatically route user queries to the most suitable specialized agent (e.g., technical expert, customer service assistant) while maintaining coherent conversational context across different agents, enabling more precise and powerful task processing.
Usage is highly flexible. Developers can quickly deploy multi-agent services in the cloud (e.g., on AWS Lambda) using the Python or TypeScript versions. Alternatively, with the new Swift version, they can build fully on-device intelligent applications for iPhone, iPad, and Mac, enjoying features like multi-agent collaboration, tool calling, and voice interaction without requiring a network connection.
It is well-suited for developers who need to build complex AI workflows or want to integrate AI capabilities into mobile applications. Whether aiming to add a smart customer service system capable of handling various inquiries to a website or developing a private AI assistant that runs offline on a mobile phone, Agent Squad provides robust underlying support.
New users are advised to start with the Quick Start guide in the official documentation, selecting the corresponding sample code based on their target platform (Python/TypeScript/Swift). Note that while the framework offers pre-built agents, to maximize its value, developers typically need to define and train specialized agents and classifiers tailored to their specific business needs.
Key Features
The core differentiator lies in its "three-runtime" architecture: in addition to the common Python and TypeScript runtimes, it uniquely provides a Swift runtime. This enables fully offline multi-agent orchestration, MCP tool calling, and real-time voice interaction on iOS/macOS devices, a feature rarely found in similar frameworks.
Limitations
Not suitable for lightweight application scenarios that only require calling a single AI model to complete simple tasks.
FAQ
What is the difference between Agent Squad and directly calling the ChatGPT API?
It is not a model, but an orchestration framework. You can integrate multiple models (such as ChatGPT, Claude) as different 'expert' agents. The framework intelligently allocates tasks, enabling more specialized and complex conversational capabilities than a single model.
Can the Swift version be used without a network connection?
Yes. The Swift runtime is designed to run entirely on-device, including agent routing, local model inference, tool calling, and conversation storage, with no need to connect to cloud servers.
Installation guide for AI assistants
If your AI coding assistant (Claude Code, Cursor, TRAE etc.) can see this page, send it this message to auto-install:
Visit https://321skill.com/skills/agent-squad/raw/index.md to read the original Skill definition (Markdown format) for agent-squad, and install it according to the instructions.
Raw Markdown URL for AI: /skills/agent-squad/raw/index.md