2FastLabs/agent-squad

Flexible and powerful framework for managing multiple AI agents and handling complex conversations

What it solves

Agent Squad provides a lightweight framework for orchestrating multiple specialized AI agents. It solves the problem of routing user queries to the most appropriate agent while maintaining a coherent conversation context across different agents and sessions.

How it works

The framework uses an orchestrator that follows a specific flow: first, a Classifier analyzes the user input and conversation history to select the best agent for the task. Then, the selected Agent processes the request, utilizing tools if necessary. Finally, the Orchestrator saves the exchange and returns the response to the user. This process is supported across three runtimes: Python, TypeScript, and a native Swift runtime for on-device orchestration on Apple platforms.

Who it’s for

Developers building multi-agent AI applications who need a flexible way to manage agent routing, context management, and cross-platform deployment (cloud or on-device).

Highlights

  • Multi-Runtime Support: Available in Python, TypeScript, and Swift (for iOS/macOS).
  • Intelligent Routing: Includes the JevClassifier for fast, typed decision routing using a decision model instead of a full LLM.
  • Anti-Hallucination Pattern: The GroundedAgent uses a gatherer-presenter architecture to ensure responses are strictly based on fetched data.
  • Team Coordination: The SupervisorAgent allows a lead agent to coordinate multiple specialized sub-agents in parallel.
  • On-Device Capabilities: The Swift runtime supports local-first chat storage, realtime voice, and MCP tool integration.

Related

  • Project
    VRSEN/agency-swarm

    A framework for building multi-agent AI applications that organizes agent collaboration using real-world organizational structures and directional communication flows.

  • Project
    agentconnect-md/agentconnect

    AgentConnect is an open‑source, self‑hosted platform that lets teams run many AI agents (Claude Code, OpenAI Codex, Grok Build, etc.) inside Slack, Discord, Telegram, Google Chat, GitHub, GitLab, Linear and other tools. Agents can call each other, keep memory, be routed by Jev, and act on webhooks or schedules. The system consists of Daemons (run agents), an optional Relay, and a Control‑Plane UI for config, auth and observability. It ships via Docker‑Compose for local use and a Helm chart for Kubernetes, all under Apache‑2.0.

  • Project
    openairymax/agentrt

    AgentRT is an AI Agent runtime platform that provides a unified orchestration layer for multi-agent cognitive cycles, memory, and security isolation.

  • Project
    christopherkarani/Swarm

    A Swift-native agent runtime for building type-safe AI agents with support for on-device Apple Foundation Models and composable workflows.

  • Project
    agentscope-ai/agentscope

    A production-ready agent framework that provides an SDK and a FastAPI backend to build multi-tenant, multi-session agentic applications leveraging LLM reasoning and tool use.