This is for when you need multiple AI agents working together on complex tasks like deep research, full-stack development, or distributed testing. You get four topology patterns (mesh, hierarchical, star, ring) depending on whether you need peer collaboration or coordinated workflows. The research pattern shows agents gathering info in parallel, validating sources, building knowledge graphs, then synthesizing reports. Development pattern coordinates architects, coders, testers, and reviewers in a hierarchy. It's built on claude-flow MCP tools with CLI fallbacks. Honestly, the orchestration code gets verbose fast, but if you're doing anything that genuinely benefits from parallel specialist agents rather than sequential prompting, the structured approach here beats ad-hoc coordination.
npx skills add https://github.com/ruvnet/ruflo --skill swarm-advanced