AGENTCONN
R

// Agent profile

Ruflo

Coding Free

About Ruflo

ruvnet/ruflo (formerly claude-flow) is an open-source agent meta-harness that adds a nervous system to Claude Code and Codex. It coordinates 100+ specialized agents via swarms, provides self-learning memory that improves across sessions, supports federated communication across machines, and ships enterprise security guardrails. Written primarily in TypeScript, Ruflo installs via a single `npx ruflo init` and operates through hooks that automatically route tasks, learn from successful patterns, and coordinate agents in the background. Available as both a full CLI install (98 agents, 60+ commands, 30 skills, MCP server, hooks, daemon) and a lighter Claude Code plugin path (35 plugins, slash commands only). Stats as of September 2026: 71.1K GitHub stars, 8.4K forks, MIT-licensed — high-velocity development with 8.1M+ ecosystem downloads. Powered by Cognitum.One agentic architecture with a Rust-based AI engine, embeddings, and plugin system.

Key Features

  • 100+ specialized agents with self-organizing swarm coordination
  • Self-learning memory that improves across sessions
  • Federation — agents on different machines communicate securely
  • Native Claude Code, Codex, and Model Context Protocol (MCP) support
  • 35 Claude Code plugins with slash command interface
  • Built-in RAG, embeddings, and knowledge graph via RuVector
  • Enterprise security guardrails and observability hooks
  • Two install paths: full CLI (npx ruflo init) or lightweight plugin

Overview

Ruflo (formerly claude-flow) is an open-source agent meta-harness from ruvnet that turns Claude Code and Codex into multi-agent systems. One npx ruflo init installs a complete coordination layer — 98 agents, 60+ CLI commands, 30 skills, an MCP server, hooks, and a background daemon — so agents self-organize into swarms, learn from every task, remember across sessions, and communicate securely across machines via federation. With 71K GitHub stars and 8.1M+ ecosystem downloads, it is among the most widely adopted agent orchestration frameworks in the Claude ecosystem.

Key Capabilities

Ruflo’s central abstraction is the swarm: a coordinated set of agents that share state, route tasks based on capability, and recover from individual agent failures without bringing down the system. The self-learning memory system means agents improve over time — successful patterns are captured and reapplied. Federation enables agents on different machines to communicate securely without leaking data. The hooks system automatically routes tasks and coordinates agents in the background, so after installation the developer keeps working normally in Claude Code while Ruflo handles orchestration. For lighter adoption, 35 Claude Code plugins provide slash commands without the full install.

Why It Matters

Most teams adopting Claude Code or Codex hit a ceiling once they need more than one agent working concurrently or coordinating handoffs. Hand-rolling that orchestration is non-trivial — task routing, shared state, partial-failure handling, and observability all become real engineering problems. Ruflo provides the orchestration scaffolding so teams can focus on the agent logic itself. The two-path install (full CLI vs lightweight plugins) means teams can adopt incrementally — start with a few slash commands and scale to the full daemon when the workflow demands it.

Use Cases

Best for teams running 3+ specialized agents that need to coordinate (research, planning, implementation, review pipelines), production systems that need observability across agent boundaries, and projects building on top of MCP that want a first-class harness rather than wiring everything by hand. The plugin marketplace (35 plugins including ruflo-swarm, ruflo-rag-memory, ruflo-neural-trader) lets teams compose exactly the capabilities they need. Often paired with skill directories like mattpocock/skills or Anthropic Skills to give individual agents in the swarm specialized capabilities.

Similar Agents