Show HN: ProxyBoy. A Windows HTTP/HTTPS debugging proxy with an AI assistant Show HN：ProxyBoy。一款带有AI助手的Windows HTTP/HTTPS调试代理

TL;DR

• ProxyBoy is an experimental, Windows-native HTTP/HTTPS debugging proxy built with Electron, designed as a lightweight alternative to mature tools like Proxyman.
• The project integrates the GitHub Copilot SDK to provide an embedded AI assistant capable of analyzing traffic, creating debugging rules, and explaining network flows conversationally.
• Key technical features include MITM traffic interception, support for GraphQL, Protobuf/gRPC, and WebSocket/SSE, along with standard proxy functionalities like breakpoints and HAR export.
• The AI agent utilizes specific tools such as getRecentTraffic, searchTraffic, and createBreakpointRule to interact with the proxy engine, allowing for automated rule creation and traffic analysis.
• Built using React, TypeScript, and sql.js for persistence, the project serves primarily as a learning exercise for the developer rather than a production-ready commercial product.

Why It Matters

This project demonstrates a practical application of embedding agentic AI capabilities directly into developer tooling, moving beyond simple code completion to active workflow automation within debugging environments. For AI practitioners, it offers a concrete example of how LLMs can be integrated with local system tools via SDKs to perform stateful operations like modifying network traffic or generating configuration rules. It highlights the emerging trend of "AI-native" developer utilities that leverage conversational interfaces to lower the barrier for complex debugging tasks.

Technical Details

• Architecture: The application is built on Electron with a React and TypeScript frontend, utilizing Electron Forge for packaging and Tailwind CSS for styling. Persistence is handled via sql.js (SQLite in-process).
• AI Integration: Powered by the @github/copilot-sdk, the embedded assistant acts as an agent with access to specific tools including getRecentTraffic, searchTraffic, getErrorFlows, getFlowDetails, createBreakpointRule, createMapLocalRule, exportHar, and controlProxy. Tool execution can be configured for auto-approval or manual confirmation.
• Proxy Engine Features: Implements Man-in-the-Middle (MITM) interception with automatic SSL certificate generation. Supports advanced protocol decoding for GraphQL, Protobuf (with .proto file support), and gRPC. Includes features for network throttling, cookie inspection, and WebSocket/SSE capture.
• Rule Management: Allows users to create Breakpoint Rules (pause/inspect/forward/drop), Map Local Rules (mock APIs with local files), and Map Remote Rules (redirect traffic). Script rules enable sandboxed JavaScript execution for request/response rewriting.
• Dependencies & Requirements: Requires Node.js 20+, Windows 10/11, and a GitHub Copilot subscription for the AI features (though the proxy functions independently). Uses react-virtuoso for efficient rendering of large traffic lists.

Industry Insight

The integration of agentic AI into debugging proxies suggests a shift towards "conversational debugging," where developers describe intent or errors in natural language rather than manually configuring filters and breakpoints. This could significantly reduce the cognitive load associated with complex network troubleshooting. However, the reliance on third-party SDKs like Copilot for core functionality raises questions about vendor lock-in and the stability of AI-dependent developer tools in enterprise environments. Additionally, the experimental nature of this project underscores the current gap between AI-enhanced prototypes and robust, security-hardened production-grade network tools.

TL;DR

• ProxyBoy 是一款基于 Electron 构建的 Windows 原生 HTTP/HTTPS 调试代理工具，旨在提供类似 Proxyman 的功能但专注于 Windows 平台。
• 核心创新在于深度集成 GitHub Copilot SDK，通过嵌入式 AI 助手实现对网络流量的自然语言查询、模式分析及自动化规则创建。
• 具备完整的中间人（MITM）代理功能，包括 SSL 证书自动管理、GraphQL/gRPC 解码、断点拦截、本地/远程映射及 HAR 导出等专业特性。
• 采用 Electron + React + TypeScript 技术栈，利用 sql.js 进行本地持久化存储，并支持沙箱化的 JavaScript 脚本规则重写。
• 该项目定位为个人实验性作品，强调通过构建此类工具来学习 AI Agent 能力与桌面应用开发的结合，而非替代成熟的商业软件。

为什么值得看

对于希望探索 AI 如何重塑传统开发者工具（如网络调试器）的从业者而言，ProxyBoy 提供了一个将 LLM Agent 能力嵌入本地工作流的优秀参考案例。它展示了如何通过 SDK 将自然语言交互转化为具体的代理操作（如创建断点、搜索流量），为“AI-Native”桌面应用的设计提供了实践思路。

技术解析

• AI 代理集成：利用 @github/copilot-sdk 实现嵌入式 AI 助手，提供 getRecentTraffic、searchTraffic、getErrorFlows 等工具函数，支持自动或手动确认执行，实现对话式流量分析与调试。
• 核心代理引擎：基于 MITM 原理捕获 HTTP/HTTPS 流量，自动生成根 CA 证书以解密 HTTPS 通信；支持 WebSocket、SSE、GraphQL 及 Protobuf/gRPC 解码，具备请求/响应拦截、修改及重定向能力。
• 前端与状态管理：使用 React 和 TypeScript 构建 UI，Tailwind CSS 负责样式，react-virtuoso 优化大量流量数据的虚拟化渲染性能，sql.js 在浏览器环境中提供 SQLite 级别的本地数据持久化。
• 构建与运行环境：基于 Electron Forge 打包，要求 Node.js 20+ 和 Windows 10/11 系统；AI 功能需 GitHub Copilot 订阅，但基础代理功能可独立运行。

行业启示

• 工具智能化趋势：传统静态功能的开发工具正通过集成 AI Agent 向动态、交互式体验转变，自然语言界面将成为复杂调试任务的新入口。
• 本地优先的 AI 应用：在桌面端本地部署 AI 能力（即使依赖云端 API）能更好地保护隐私并利用系统级权限（如网络拦截），为敏感数据场景下的 AI 集成提供范式。
• 开源实验的价值：此类实验性项目虽不成熟，但快速验证了特定技术组合（如 Electron + Copilot SDK）的可行性，有助于行业探索下一代开发者工具的技术栈选型。