RuFlo
Agent-orchestration platform for Anthropic Claude (Claude Code and Codex) with multi-agent swarms, federation, HNSW vector memory via AgentDB, ~100 specialized agents, ~210 MCP tools, and 33 plugins. Renamed from claude-flow at v3.5.0 in early 2026; current version v3.10.5.
Overview
RuFlo (renamed from claude-flow at v3.5.0 in early 2026; current version v3.10.5) is my go-to agent-orchestration platform for complex AI development workflows. It sits on top of Claude Code / Codex and coordinates fleets of specialized agents via swarm topologies, SPARC methodology, hook automation, persistent HNSW-indexed vector memory, and tight GitHub integration. The platform ships with ~100 specialized agents, ~210 MCP tools, and 33 plugins.
How I Use RuFlo
I leverage RuFlo for full-stack development, orchestrating specialized agents for research, coding, testing, and review. The SPARC methodology ensures systematic development with proper architecture and testing.
Agent Specialization
I configure swarms with specialized agents: researchers for requirements analysis, coders for implementation, testers for quality assurance, and reviewers for code quality. Each agent has specific capabilities and coordinates through shared memory.
Automation Benefits
Hooks automatically format code, validate changes, and update memory. This automation reduces cognitive load and ensures consistency across projects.
SPARC Methodology
SPARC provides a systematic approach to development: Specification (requirements), Pseudocode (algorithm design), Architecture (system design), Refinement (TDD implementation), and Completion (integration). This methodology ensures thorough planning and quality.
Memory & Federation
AgentDB provides HNSW-indexed vector memory with 150x faster semantic retrieval, enabling context retention across sessions. Federation connects installations across operators for team-scale orchestration. Cross-session memory lets each session pick up where the last left off without re-ingesting the entire repo.