[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$f2i6vpksbqo14p":3},{"code":4,"message":5,"data":6,"timestamp":57},10000,"success",{"full_name":7,"position":8,"description":9,"language":10,"stars":11,"forks":12,"stars_gained":13,"age_days":14,"hn_threads":15,"one_liner":27,"summary_md":28,"topic":29,"github_topics":32,"license":53,"homepage":54,"updated_at":55,"history":56},"ruvnet\u002Fruflo",2,"🌊 The original agent harness. Deploy intelligent multi-player swarms, coordinate autonomous workflows, and build conversational AI systems. Features adaptive memory, self-learning intelligence, federation, vector RAG integration, and native Claude Code \u002F Codex \u002F Hermes and many more Integrated","TypeScript",78966,9273,3349,508,[16,22],{"id":17,"title":18,"points":19,"comments":20,"posted_at":21},49051072,"Ruflo: An agent meta-harness for Claude Code and Codex",3,0,"2026-07-25",{"id":23,"title":24,"points":25,"comments":25,"posted_at":26},47943679,"Is any one using ruflo?",1,"2026-04-29","给 Claude Code 和 Codex 套上多智能体集群、持久记忆与跨机联邦的运行时外壳。","## 解决什么问题\n\nClaude Code 和 Codex 本质是单智能体循环：一个模型、一个上下文窗口、一台机器。任务一复杂，你就得手工拆子任务、在会话之间复制上下文，上一轮学到的东西下一轮全丢。Ruflo 自称\"agent meta-harness\"，把协调、记忆和路由包在这些编码智能体外面，让多个 agent 协作，并把经验沉淀下来。\n\n## 有什么不同\n\n它没有另起炉灶造一个 agent 框架（像 LangGraph、CrewAI 这类编排库），而是寄生在现有宿主上：Claude Code 通过插件市场接入，其他支持 MCP 的客户端连本地 stdio server。真正拉开差距的是三件事：让专用 agent 自组织的 swarm 编排、把历史任务结果回灌到路由的向量记忆（RuVector）、以及跨机器通信的 federation。这和纯库的取舍不同——你继承宿主的权限和工具链，而不是替换它。\n\n## 适合谁\n\n最适合已经长期泡在 Claude Code 或 Codex 里、想要并行拆任务、跨会话记忆和可视化控制台又不想换编辑器的开发者。跨多台机器做多 agent 实验的团队是 federation 的目标用户。如果你要的是一个精简可审计的 agent 循环，或者并不打算绑定 Claude Code\u002FCodex，那就该去看普通编排库——这里的价值在宿主集成，而它的表面积非常大。\n\n## 上手成本\n\n插件路线要求 Claude Code 2.1.287 及以上；其他客户端需要配置 MCP，例如 `npx ruflo@latest init wizard`。MIT 协议，TypeScript\u002FNode 实现，npm 工具链是前提。README 声称 hooks 能自动路由和学习，但同时明确说 mods 以你账号权限运行、安装前要自己审代码——凡是要碰你仓库的东西，这都是实打实的风险。README 塞了大量营销内容（花哨动图、ruOS 推广、\"示意性下载计数器\"），学习循环也被标注为概念示意而非实时执行轨迹；能力宣称最好自己在本项目里验证后再信。\n\n## 点评\n\n从仓库年龄看（2025 年 6 月创建，约 500 天）这像个成熟项目，星数接近 7.9 万也相当夸张，最近一周以约 3300 星冲到周榜第 4。但它从未上过日榜，Hacker News 上几乎没有讨论（2 分、0 评论），星数与可见的社区声量明显不匹配。结论是：有真实沿革和真实产品面，但当前这波热度值得打个问号。先在临时仓库里试，别一上来就压上主力工作流。",{"slug":30,"name":31},"ai-engineering","AI 工程",[33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52],"agentic-ai","agentic-framework","agentic-workflow","agents","ai-agents","ai-assistant","ai-skills","autonomous-agents","claude","codex","dsh-plugin","harness","mcp-server","multi-agent","multi-agent-systems","npm","skills","swarm","swarm-intelligence","typescript","MIT","https:\u002F\u002FCognitum.One","2026-10-24T16:35:44Z",[],1792868584]