[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$f12lzu99izt5gx":3},{"code":4,"message":5,"data":6,"timestamp":72},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":36,"summary_md":37,"topic":38,"github_topics":41,"license":62,"homepage":63,"updated_at":64,"history":65},"ZSeven-W\u002Fopenpencil",12,"The world's first open-source AI-native vector design tool and the first to feature concurrent Agent Teams. Design-as-Code. Turn prompts into UI directly on the live canvas. A modern alternative to Pencil.","Rust",6594,576,100,249,[16,22,28,32],{"id":17,"title":18,"points":19,"comments":20,"posted_at":21},47493551,"OpenPencil: Open-source AI-native vector design tool",5,0,"2026-03-23",{"id":23,"title":24,"points":25,"comments":26,"posted_at":27},47151317,"OpenPencil: Open-source vector design tool controlled by AI Agents",4,2,"2026-02-25",{"id":29,"title":30,"points":26,"comments":26,"posted_at":31},47195713,"Show HN: OpenPencil - Open-source vector design tool controlled by AI Agents","2026-02-28",{"id":33,"title":34,"points":35,"comments":20,"posted_at":27},47147015,"Show HN: OpenPencil – An open-source, AI-native vector design tool",1,"把提示词和并行 Agent 直接变成画布上的 UI，并导出为代码的 AI 原生矢量设计工具。","## 解决什么问题\n\n用矢量工具做 UI 设计是典型的手工循环：先画 frame，再交给前端重写一遍同样的布局。市面上给设计工具加的 AI 插件，大多只生成一张静态图或粗略稿，交付环节照旧。OpenPencil 想砍掉这一环：它把设计文件本身做成可读、可 diff、可导出代码的文本产物，并让 AI Agent 在实时画布上生成和修改。\n\n## 有什么不同\n\n现有 AI 设计产品多数是挂在成熟编辑器上的插件，产出是图片或结构松散的图层。OpenPencil 从自己的 `.op` 格式出发——JSON 文件，可读可 diff，能转成 React + Tailwind、HTML + CSS、Vue、Svelte、Flutter 以及原生 UI 目标，所以产出更接近代码而不是一张图。真正有辨识度的是两点：并发 Agent Teams，由 orchestrator 把页面拆成空间子任务，多个 Agent 并行处理 hero、features、footer 等区域，并在画布上逐个显示成员状态；以及内置 MCP Server，让 Claude Code、Codex 这类终端 Agent 直接读写 `.op` 文件。多模型处理也不是假设单一后端，而是按模型能力调整提示词。代价是：你要接受一个年轻的格式和生态，而不是已有的成熟生态。\n\n## 适合谁\n\n适合希望设计稿能直接导出真实组件的前端和产品工程师，也适合愿意 AI-first 而非迁就现有 Figma 库的小团队；习惯在终端里、用支持 MCP 的 Agent 驱动设计而不是拖鼠标的开发者会更顺手。如果你需要成熟的插件生态、企业级设计系统治理，或者与专职设计团队做像素级交付，那这里还不合适——它目前不是 Figma 的替代品。\n\n## 上手成本\n\nRust 编写，以原生二进制分发：Homebrew、Scoop、GitHub Releases 直装，以及 Nix（flake 目前只发布 Linux x86_64）。MIT 协议，商用没有额外限制。但它依赖外部模型 API，README 里的赞助横幅指向第三方 API 转售服务，意味着你要自备 key、按 token 付费，输出质量取决于你配置的模型。README 功能与演示写得很足，局限性几乎没提，并专门说明了 Windows 安装包被杀的已知误报。协作走带公共 relay 的点对点会话。macOS 上通过 `brew tap zseven-w\u002Fopenpencil` 和 `brew install --cask openpencil` 安装。\n\n## 点评\n\n这是一个很年轻的项目：创建于 2026 年 2 月，至最新信号约 249 天，热度呈短促爆发——约 6.5k stars，日榜只上过一天（+103），Rust 周榜一周 +315。这更像一个抓眼球的 demo 带来的早期尖峰，而不是已经站住脚的工具。Agent Teams、MCP 原生设计文件、代码导出这几个方向是自洽的，MIT 协议也让试错成本不高；但 2026 年初才起步的代码库、公开记录还很薄，要预期粗糙边缘和格式变动。对既是开发者又做设计的人，值得花一个下午；现在还不适合团队标准化采用。",{"slug":39,"name":40},"media","多媒体与设计",[42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61],"agent","agent-team","ai","claude","claude-code","codex","dsh-plugin","fimga","flutter","mcp","opencode","openpencil","pencil","react","react-native","rust","skill","ui","vibecoding","vibedesign","MIT","https:\u002F\u002Fop.zseven.tech","2026-10-24T23:05:48Z",[66,70],{"date":67,"stars":68,"stars_gained":69},"2026-10-24",6577,103,{"date":71,"stars":11,"stars_gained":13},"2026-10-25",1792898609]