What is Ruflo?
Ruflo is the #1 ranked project on GitHub right now. It's completely free, open source, and it deploys 60+ AI agents that work together, share memory, and get smarter after every run. The full setup guide is below ā it's easier than you think.
ā GitHub Repo: github.com/ruvnet/ruflo
š Lost Before You Even Started?
ā” Before You Run a Single Command
No terminal. No Claude Code. No idea what npm install does ā and that's completely fine.
That's exactly what The AI Surfer is built for.
We train people to become Agent Commanders ā people who don't just use AI tools, they deploy and direct them. Terminal basics, Claude Code setup, multi-agent swarms ā all of it, step by step, from zero.
Once you have the foundation, setting this up takes under 5 minutes.
ā Join us at theaisurfer.com and become an Agent Commander.
Install Ruflo (Pick One ā All Work)
Option 1: Claude Code Plugin (Recommended if you're in Claude Code)
Paste these directly into your Claude Code session:
/plugin marketplace add ruvnet/ruflo
/plugin install ruflo-core@ruflo
/plugin install ruflo-swarm@ruflo
/plugin install ruflo-autopilot@ruflo
Option 2: One-Line Install (Terminal)
curl -fsSL https://cdn.jsdelivr.net/gh/ruvnet/ruflo@main/scripts/install.sh | bash
Option 3: NPX (Quickest way to try it)
npx ruflo@latest init --wizard
Option 4: Global Install
npm install -g ruflo@latest
Option 5: Add as MCP Server in Claude Code
claude mcp add ruflo -- npx -y @claude-flow/cli@latest
After Install ā First Run Best Practices
- ā
Run the wizard first ā
npx ruflo@latest init --wizardwalks you through your first swarm setup - ā Start with a real task ā don't test with toy prompts. Give it something with actual complexity (a full feature, a codebase audit, a multi-step research task)
- ā Let it run a few times ā the collective memory compounds. Run 1 is good. Run 5 is significantly better.
- ā Check your API keys are set ā Ruflo routes between models automatically. Make sure your Claude/OpenAI keys are configured so the routing actually works
- ā
Install ruflo-intelligence for the self-learning layer:
/plugin install ruflo-intelligence@ruflo - ā
Install ruflo-cost-tracker to monitor token usage:
/plugin install ruflo-cost-tracker@ruflo - ā ļø Don't install all 32 plugins at once ā start with core + swarm, add as needed
What's Actually Happening When You Run It
- Queen agents manage the teams and break your task into workstreams
- 8 tactical worker types handle execution: Researcher, Coder, Analyst, Tester, Architect, Reviewer, Optimizer, Documenter
- 5 consensus protocols (Raft, BFT, Gossip, CRDT) ā agents check each other's work before output moves forward
- Shared collective memory ā every agent writes back what it learns. The swarm gets smarter across every session
- Automatic model routing ā simple tasks go to cheap models, complex ones escalate. Up to 50% token reduction, up to 250% more effective Claude usage
You're Now Ready
Paste the commands, run the wizard, give it a real task, and let the swarm work.
The repo: github.com/ruvnet/ruflo The community: theaisurfer.com