squad  by mco-org

Terminal-based multi-AI agent collaboration

Created 3 months ago
558 stars

Top 56.7% on SourcePulse

GitHubView on GitHub
Project Summary

Multi-AI agent terminal collaboration tool that enables multiple AI CLI agents to communicate and work together via simple shell commands and an SQLite database. It targets users of AI CLI tools like Claude, Gemini, and Codex, offering real-time, daemon-less collaboration for complex tasks by assigning roles such as manager, worker, and inspector across different terminals.

How It Works

Squad facilitates agent communication through a shared SQLite database (.squad/messages.db), eliminating the need for daemons or background processes. Each agent operates within its own terminal, using CLI commands (squad send, squad receive, squad task) to exchange messages and tasks. This approach ensures each operation is a one-shot execution, simplifying debugging and providing a clear, traceable message flow managed via structured task assignments or freeform coordination.

Quick Start & Requirements

  • Installation: macOS: brew install mco-org/tap/squad. Windows: Download prebuilt binary from GitHub Releases. Build from source: cargo install --git https://github.com/mco-org/squad.git.
  • Prerequisites: Rust 1.77+ (for building). The optional tmux launcher requires tmux and ruby. Integration with AI tools requires the respective CLI tools (Claude Code, Gemini CLI, Codex CLI, OpenCode) and running squad setup to register the /squad command.
  • Setup: Run squad init in a project directory to initialize the workspace and .squad/ configuration.
  • Links: GitHub Releases: https://github.com/mco-org/squad/releases

Highlighted Details

  • Real-time, multi-agent collaboration managed via distinct roles (manager, worker, inspector) in separate terminals.
  • SQLite serves as a central, persistent message bus, ensuring no daemons or background processes are required.
  • Automatic ID suffixing handles simultaneous joins of agents with identical role names (e.g., worker, worker-2).
  • An optional tmux launcher script automates the setup of multi-terminal agent sessions for specific workflows.
  • Customizable agent roles can be defined via .md files in .squad/roles/.

Maintenance & Community

No specific details regarding maintainers, community channels (like Discord/Slack), or roadmap were present in the provided README.

Licensing & Compatibility

Licensed under the MIT License, permitting commercial use and integration into closed-source projects.

Limitations & Caveats

The optional tmux launcher has specific dependencies (tmux, ruby, claude) and is maintained separately from the core Rust CLI. Integration relies on the specific AI CLI tools and their command registration mechanisms. While the SQLite approach simplifies architecture, managing complex inter-agent dependencies and ensuring robust error handling across many agents might require careful design.

Health Check
Last Commit

1 month ago

Responsiveness

Inactive

Pull Requests (30d)
0
Issues (30d)
0
Star History
434 stars in the last 30 days

Explore Similar Projects

Starred by Andrew Ng Andrew Ng(Founder of DeepLearning.AI; Cofounder of Coursera; Professor at Stanford), Jack Lukic Jack Lukic(Author of Semantic UI), and
5 more.

ag2 by ag2ai

0.4%
5k
AgentOS for building AI agents and facilitating multi-agent cooperation
Created 1 year ago
Updated 11 hours ago
Feedback? Help us improve.