Plan a safe first run with an AI agent swarm
Ruflo turns Claude Code into a team of specialised agents with a shared memory. Here's what it does, and a prompt to plan a safe first run before you install it.
By Alan Brown · Works with Claude Code
🐝 This free GitHub tool runs 60+ AI agents as one team.
- 🤖 It's called Ruflo. Instead of one AI agent doing everything, it spins up a swarm of specialised agents. Queen agents coordinate the teams, and worker agents do the research, coding, testing and reviewing.
- 🧠 They all share one memory, so the system learns from what worked and gets better over time.
- 🔀 The part that got me: it reads how complex your task is and routes it to a suitable model. Simple tasks go to cheaper models, and heavy builds go to the bigger ones.
- 🔓 It's open source (MIT) and plugs straight into Claude Code. github.com/ruvnet/ruflo
Context & limitations
The README lists 100+ agents and says its results depend on how you configure memory and learning, so "gets better over time" is the project's claim, not a guarantee. The full install adds agents, hooks and settings to your project folder, so try it on a branch first.
Do this yourself
A swarm makes lots of changes quickly, so start with one small, checkable task on a branch rather than your whole project. Ruflo works with Claude Code and Codex; the planning prompt works in any AI.
Want to build this with a teacher?
Pathwise Academy runs small, hands-on AI classes in Singapore for ages 10–17 and adults.

