Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/ruvnet/ruflo/llms.txt

Use this file to discover all available pages before exploring further.

Agent = Model + Harness. The model writes; the harness gives it tools, memory, loops, sandboxes, and controls so it can actually work. Ruflo is the harness — the execution layer around Claude Code and Codex that adds 100+ specialized agents, coordinated swarms, self-learning memory, federated communication across machines, and enterprise security guardrails. So agents don’t just run; they collaborate, learn from every task, remember across sessions, and — with federation — securely talk to agents on other machines without leaking data. One npx ruflo@latest init gives Claude Code a nervous system. You keep writing code. Ruflo handles the coordination.
New users don’t need to learn all 313 tools. After init, just use Claude Code normally — the hooks system automatically routes tasks to the right agents, learns from successful patterns, and coordinates multi-agent work in the background. The advanced tools exist for fine-grained control when you need it.

​
Key Capabilities

100+ Specialized Agents

Ready-to-use agents for coding, code review, testing, security audits, documentation, and DevOps. Each agent is optimized for its specific role across 8 categories including core development, SPARC methodology, GitHub automation, and swarm coordination.

Swarm Coordination

Run unlimited agents simultaneously in organized swarms with hierarchical, mesh, ring, and star topologies. Agents spawn sub-workers, share context, and divide work automatically using queen-led hierarchy with 5 consensus algorithms: Byzantine, Raft, Gossip, CRDT, and Quorum.

Self-Learning Memory

SONA self-learning with sub-millisecond pattern matching, HNSW-indexed AgentDB vector memory, and EWC++ to prevent catastrophic forgetting. The system remembers what works and routes similar tasks to the best-performing agents — getting smarter with every session.

313 MCP Tools

Native Model Context Protocol integration with 313 tools across 31 modules: coordination, memory, neural learning, GitHub operations, background workers, and lifecycle hooks. All tools are available directly from Claude Code once the MCP server is registered.

27 Hooks + 12 Workers

27 lifecycle hooks (pre/post-edit, pre/post-command, session, intelligence, and more) plus 12 auto-triggered background workers for auditing, optimization, test-gap detection, documentation, and memory consolidation — all firing automatically as you work.

35 Plugins + 137+ Skills

A full plugin marketplace with 35 native Claude Code plugins covering every domain: memory, swarm coordination, security, RAG retrieval, browser testing, ADR tracking, DDD scaffolding, IoT, and more. Plus 137+ reusable skills for Codex and Claude Code.

​
Architecture

Ruflo places a self-learning orchestration layer between you and the model. Every request flows through intelligent routing, reaches specialized agents, feeds results back into memory, and improves future routing — a continuous learning loop: 
Self-Learning / Self-Optimizing Agent Architecture

User --> Ruflo (CLI/MCP) --> Router --> Swarm --> Agents --> Memory --> LLM Providers
                          ^                           |
                          +---- Learning Loop <-------+
Expanded, the full system looks like this: 
User → Claude Code / CLI
          |
          v
    Orchestration Layer
    (MCP Server, Router, 27 Hooks)
          |
          v
    Swarm Coordination
    (Queen, Topology, Consensus)
          |
          v
    100+ Specialized Agents
    (coder, tester, reviewer, architect, security...)
          |
          v
    Memory & Learning
    (AgentDB, HNSW, SONA, ReasoningBank)
          |
          v
    LLM Providers
    (Claude, GPT, Gemini, Cohere, Ollama)
The Learning Loop is not optional — after each task completes, the hooks system scores outcomes, stores successful patterns in vector memory (HNSW-indexed, sub-millisecond retrieval), and updates the router so future similar tasks go to the best-performing agents automatically.

​
Two Install Paths

There are two distinct ways to use Ruflo with very different surface areas. Choose based on what you need:
Claude Code PluginCLI Install (npx ruflo@latest init)
What it gives youSlash commands + skills + agent definitions per pluginFull Ruflo loop — 100+ agents, 60+ commands, MCP server, hooks, daemon
Files in your workspaceZero.claude/, .claude-flow/, CLAUDE.md, helpers, settings
MCP server registeredNoYes
memory_store, swarm_init availableNoYes
Hooks installedNoYes
Best forTry a single plugin’s commands without committing to the full installProduction use — everything works as documented

​
Claude Code Plugin (Lite)

Installing via the Claude Code plugin marketplace gives you slash commands and agent definitions only. The Ruflo MCP server is not registered in this mode, so tools like memory_store, swarm_init, agent_spawn, and hooks_route will not be callable from Claude Code sessions. This path is ideal for evaluating specific plugin capabilities before committing to the full setup. 
/plugin marketplace add ruvnet/ruflo
/plugin install ruflo-core@ruflo
/plugin install ruflo-swarm@ruflo

​
CLI Install (Full Loop)

Running npx ruflo@latest init installs the complete Ruflo system: 100+ agents, 26 CLI commands with 140+ subcommands, the full MCP server with 313 tools, 27 hooks, 12 background workers, 35 plugins, and 137+ skills. This is the path for production use where everything in the documentation applies. 
npx ruflo@latest init wizard   # Interactive setup
npx ruflo@latest init          # Non-interactive
After init, register the MCP server with Claude Code and you’re done: 
claude mcp add ruflo -- npx ruflo@latest mcp start

​
What’s in the Box

After a full CLI install, Ruflo delivers:
CapabilityCountDescription
Specialized Agents100+Coding, testing, security, docs, architecture, GitHub, SPARC
CLI Commands26init, agent, swarm, memory, mcp, hive-mind, hooks, and more
CLI Subcommands140+Full lifecycle management for every subsystem
MCP Tools313Coordination, memory, neural, GitHub, workers, hooks
Lifecycle Hooks27Pre/post-edit, route, session, intelligence, worker, progress
Background Workers12Audit, optimize, testgaps, ultralearn, document, benchmark
Plugins35Native Claude Code plugins across all domains
Skills137+Reusable skills for Claude Code and Codex
LLM Providers5Anthropic, OpenAI, Google, Cohere, Ollama with automatic failover

​
Get Started

Quickstart

Initialize Ruflo, register the MCP server, and launch your first agent swarm in under 5 minutes.

Installation

Full install guide covering macOS, Linux, Windows (PowerShell and Git-Bash), and WSL.

Core Concepts

Deep dive into the orchestration layer, swarm topologies, memory architecture, and learning loop.

CLI Reference

Complete reference for all 26 commands and 140+ subcommands, starting with init.

Build docs developers (and LLMs) love

Get started for freeTalk to us