Summary

SwarmForge is a disciplined, tmux-based platform orchestrating AI agents as a professional software engineering team. It solves inter-agent communication and collaboration across git worktrees via structured prompts, worktree assignments, tmux sessions, and message passing. This enables reliable swarm operation on shared projects, enhancing productivity.

How It Works

This lightweight, self-hosted layer uses tmux and git worktrees. Its config-driven design defines swarm topology and roles via project-local files (swarmforge.conf, <role>.prompt). A layered constitution system allows modular prompt definition by delegating from constitution.prompt to subordinate files. Each role runs in its own tmux session/terminal window within dedicated git worktrees (.worktrees/). Agent communication uses helper scripts (notify-agent.sh). This offers flexible swarm shapes, role definitions, and real-time observability.

Quick Start & Requirements

Clone the repo, make swarmforge.sh executable, add to PATH. In your project, create swarmforge/ with swarmforge.conf and role prompts. Run swarmforge.sh <project-directory>. Requires tmux and compatible shell. No specific Python, GPU, or dataset prerequisites mentioned.

Highlighted Details

• Config-Driven Topology: Swarm structure via swarmforge/swarmforge.conf.
• Project-Local Roles: Agent behavior in dedicated prompt files.
• Layered Constitution: Modular prompts via subordinate constitution/ files.
• Per-Role Backend Selection: Supports claude, codex, or none.
• Observable Swarm: Real-time visibility via terminal windows.
• Self-Hosted & Lightweight: Operates locally with minimal resources.

Maintenance & Community

The README lacks specific details on maintainers, community channels, sponsorships, or roadmap.

Licensing & Compatibility

The README does not specify the software license, precluding assessment of commercial use or closed-source linking compatibility.

Limitations & Caveats

Operation is tied to tmux, potentially limiting use on non-tmux systems or for unfamiliar users. Setup requires strict adherence to directory structure and configuration formats. AI agent effectiveness depends on chosen backends and their APIs.