RuFlow plugs 60 coordinated AI agents into Claude Code

My notes

Summary

• RuFlow (likely “RooFlow” / “Roo Flow”) is a free open-source orchestration layer that plugs into Claude Code and spins up 60+ specialised AI agents working in coordination
• Hierarchy: “queen” agents manage teams of tactical agents (research, code, test, review) sharing collective memory that improves across runs
• Auto-routes tasks by complexity to cheap vs powerful models, claimed to cut token usage ~50% and extend Claude Code usage 2.5x

Key Insight

• The non-obvious bit is the cost-routing layer, not the agent count. Most multi-agent demos look impressive but burn tokens, RuFlow’s pitch is that it triages task complexity before picking a model
• Claim “60+ agents” sounds gimmicky; the real architecture matters more, queen/tactical split mirrors patterns like CrewAI hierarchies and AutoGen’s group chat. The differentiator is the routing + shared memory persistence
• “DM Flow” is creator promo bait, actual repo name needs verification (likely “RooFlow” on GitHub, not “RuFlow”). Always verify before installing
• Claimed 50% token reduction + 250% usage extension is the kind of metric that needs a real benchmark on production-style work before trusting. Demo creators conflate “test scenario savings” with “your savings”
• Worth testing against an existing Claude Code workflow but skeptical: the more agents you add, the more orchestration overhead and failure modes. Single Opus often beats 60 Haiku.