# 开发者工具与 AI 编程 编码 Agent、开发辅助、测试、沙箱、上下文工程、代码审查和开发者工作流。 数量:56 | 名称 | 概括 | 标签 | 链接 | | --- | --- | --- | --- | | 9router - 免费 AI 编码路由器,连接多种编码工具到免费 LLM | 免费 AI 编码路由器,连接多种编码工具到免费 LLM | tool, llm, 拿来就用 | [打开](https://github.com/decolua/9router) | | acpx - ACP协议无头CLI客户端,用于有状态Agent会话 | ACP协议无头CLI客户端,用于有状态Agent会话 | agent, multi-agent, 构建Agent, 拿来就用 | [打开](https://github.com/openclaw/acpx) | | adamsreview - 面向Claude Code的多Agent PR代码审查工具,提升代码审查质量 | 面向Claude Code的多Agent PR代码审查工具,提升代码审查质量 | multi-agent, agent, tool, 自动化 | [打开](https://github.com/adamjgmiller/adamsreview) | | aerf - 为 AI Agent 操作提供加密签名的审计收据 | 为 AI Agent 操作提供加密签名的审计收据 | multi-agent, agent, 构建Agent, 自动化 | [打开](https://github.com/aerf-spec/aerf) | | Agent - 原生 macOS AI 编码 IDE/Agent 框架,充分利用 macOS 系统能力 | 原生 macOS AI 编码 IDE/Agent 框架,充分利用 macOS 系统能力 | multi-agent, agent, framework, 构建Agent, 拿来就用 | [打开](https://github.com/macOS26/Agent) | | agent-postmortem-skill - AI编码Agent的验尸技能,强制Agent证明其工作成果 | AI编码Agent的验尸技能,强制Agent证明其工作成果 | multi-agent, agent, 构建Agent, 自动化 | [打开](https://github.com/plus8bit/agent-postmortem-skill) | | agent-sanbox - 隔离 Docker 容器内运行 AI 编码 Agent | 隔离 Docker 容器内运行 AI 编码 Agent | multi-agent, agent, 构建Agent, 部署运维 | [打开](https://github.com/marvincaspar/agent-sanbox) | | agentwrit - Go 语言的 AI Agent 凭证代理,任务级作用域凭证管理 | Go 语言的 AI Agent 凭证代理,任务级作用域凭证管理 | multi-agent, agent, 构建Agent, 部署运维 | [打开](https://github.com/devonartis/agentwrit) | | akmon - 用Rust编写的AI编码Agent,面向受监管的工程场景 | 用Rust编写的AI编码Agent,面向受监管的工程场景 | multi-agent, agent, 构建Agent, 拿来就用 | [打开](https://github.com/radotsvetkov/akmon) | | Archon - 首个开源AI编码Harness构建器,使AI编码变得确定性和可重复 | 首个开源AI编码Harness构建器,使AI编码变得确定性和可重复 | open-source, 构建Agent, 需要开发 | [打开](https://github.com/coleam00/Archon) | | BMAD-METHOD - 粒子加速器仿真 | 高级粒子加速器物理仿真库和工具包 | physics, particle-accelerator, simulation, research, scientific | - | | browsercode - 通过WebAssembly在浏览器中运行Claude Code | 通过WebAssembly在浏览器中运行Claude Code | coding, 拿来就用 | [打开](https://github.com/leaningtech/browsercode) | | chrome-devtools-mcp - Chrome DevTools MCP 服务,为编程 Agent 提供浏览器调试能力 | Chrome DevTools MCP 服务,为编程 Agent 提供浏览器调试能力 | mcp, multi-agent, tool, agent, coding, 需要开发, 构建Agent | [打开](https://github.com/ChromeDevTools/chrome-devtools-mcp) | | claude-agent-sdk-python - Anthropic 官方 Claude Agent SDK Python 版,用于构建 Claude 驱动的 AI Agent | Anthropic 官方 Claude Agent SDK Python 版,用于构建 Claude 驱动的 AI Agent | multi-agent, agent, 构建Agent, 需要开发 | [打开](https://github.com/anthropics/claude-agent-sdk-python) | | cline - IDE内自主编程Agent,支持文件创建编辑、命令执行、浏览器操作 | IDE内自主编程Agent,支持文件创建编辑、命令执行、浏览器操作 | coding, multi-agent, agent, 构建Agent, 拿来就用 | [打开](https://github.com/cline/cline) | | CloakBrowser - 隐身 Chromium 浏览器,通过所有机器人检测测试,可作为 Playwright 的直接替代品 | 隐身 Chromium 浏览器,通过所有机器人检测测试,可作为 Playwright 的直接替代品 | ai, 拿来就用, 自动化 | [打开](https://github.com/CloakHQ/CloakBrowser) | | code-review-graph - 为 Claude Code 构建本地知识图谱,代码审查 token 消耗降低 6.8 倍 | 为 Claude Code 构建本地知识图谱,代码审查 token 消耗降低 6.8 倍 | coding, 自动化, 需要开发 | [打开](https://github.com/tirth8205/code-review-graph) | | codegraph - 为 Claude Code 预索引的代码知识图谱,减少 Token 和工具调用,100% 本地运行 | 为 Claude Code 预索引的代码知识图谱,减少 Token 和工具调用,100% 本地运行 | coding, tool, 拿来就用, 自动化 | [打开](https://github.com/colbymchenry/codegraph) | | context-mode - AI编码代理上下文窗口优化,98%缩减率 | AI编码代理上下文窗口优化,98%缩减率 | ai, 需要开发, 构建Agent | [打开](https://github.com/mksglu/context-mode) | | CopilotKit - Agent 前端技术栈和 AG-UI 协议,支持 React + Angular,3 万 Star | Agent 前端技术栈和 AG-UI 协议,支持 React + Angular,3 万 Star | multi-agent, agent, 构建Agent, 需要开发 | [打开](https://github.com/CopilotKit/CopilotKit) | | crush - CharmBracelet 出品的 AI 编码 Agent | CharmBracelet 出品的 AI 编码 Agent | agent, multi-agent, 构建Agent, 拿来就用 | [打开](https://github.com/charmbracelet/crush) | | CubeSandbox - 开源 Agent 沙箱,延迟 <60ms | 开源 Agent 沙箱,延迟 <60ms | multi-agent, agent, open-source, 需要开发, 构建Agent | [打开](https://github.com/Tencent/CubeSandbox) | | DeepCode - 开放编码 Agent (Paper2Code & Text2Web & Text2Backend) | 开放编码 Agent (Paper2Code & Text2Web & Text2Backend) | agent, coding, multi-agent, 构建Agent, 拿来就用 | [打开](https://github.com/HKUDS/DeepCode) | | E2B - 开源安全沙箱执行环境,为 AI agent 提供真实世界工具,5k Star | 开源安全沙箱执行环境,为 AI agent 提供真实世界工具,5k Star | multi-agent, agent, open-source, tool, 需要开发, 构建Agent | [打开](https://github.com/e2b-dev/E2B) | | gpt-engineer - 自主代码生成Agent,开创描述需求获取代码范式 | 自主代码生成Agent,开创描述需求获取代码范式 | agent, multi-agent, 构建Agent, 拿来就用 | [打开](https://github.com/AntonOsika/gpt-engineer) | | gstack - Garry Tan的Claude Code配置:23个工具充当CEO/设计师/工程经理等角色 | Garry Tan的Claude Code配置:23个工具充当CEO/设计师/工程经理等角色 | tool, 构建Agent, 学习参考 | [打开](https://github.com/garrytan/gstack) | | InsForge - 基于 Postgres 的后端平台,为编程 Agent 设计,含认证/存储/AI 网关 | 基于 Postgres 的后端平台,为编程 Agent 设计,含认证/存储/AI 网关 | coding, multi-agent, agent, 需要开发, 构建Agent | [打开](https://github.com/InsForge/InsForge) | | ios-simulator-skill - Claude Code 的 iOS 模拟器技能,优化 Claude 构建和测试 iOS 应用的能力 | Claude Code 的 iOS 模拟器技能,优化 Claude 构建和测试 iOS 应用的能力 | ai, 拿来就用, 自动化 | [打开](https://github.com/conorluddy/ios-simulator-skill) | | jcode - 专注于多会话工作流、无限定制性和性能的 Coding Agent Harness | 专注于多会话工作流、无限定制性和性能的 Coding Agent Harness。Rust 编写。 | multi-agent, coding, agent, 构建Agent, 拿来就用 | [打开](https://github.com/1jehuang/jcode) | | kiro-gateway - Kiro IDE/CLI的代理API网关,可使用免费Claude模型 | Kiro IDE/CLI的代理API网关,可使用免费Claude模型 | ai, 拿来就用, 部署运维 | [打开](https://github.com/jwadow/kiro-gateway) | | Martin-Loop - AI编码Agent的预算上限和审计追踪工具 | AI编码Agent的预算上限和审计追踪工具 | multi-agent, agent, tool, 自动化, 构建Agent | [打开](https://github.com/Keesan12/Martin-Loop) | | mlx-code - 基于MLX的Mac本地AI编码Agent,可在后院服务器上运行 | 基于MLX的Mac本地AI编码Agent,可在后院服务器上运行 | multi-agent, agent, coding, 构建Agent, 拿来就用 | [打开](https://github.com/JosefAlbers/mlx-code) | | open-swe - LangChain 出品的异步编码 Agent | LangChain 出品的异步编码 Agent | agent, multi-agent, 构建Agent, 拿来就用 | [打开](https://github.com/langchain-ai/open-swe) | | OpenCLI - 将任何网站和工具变为 CLI,AI 原生运行时 | 将任何网站和工具变为 CLI,AI 原生运行时 | tool, 需要开发, 自动化 | [打开](https://github.com/jackwener/OpenCLI) | | opencode - 开源 AI 编码 Agent | 开源 AI 编码 Agent。 | open-source, multi-agent, coding, agent, 构建Agent, 拿来就用 | [打开](https://github.com/anomalyco/opencode) | | opengravity - 零安装BYOK的纯JS IDE克隆,AI Agent可在浏览器内Linux环境中操作 | 零安装BYOK的纯JS IDE克隆,AI Agent可在浏览器内Linux环境中操作 | multi-agent, agent, 构建Agent, 拿来就用 | [打开](https://github.com/ab-613/opengravity) | | OpenHands - AI 驱动的软件开发平台 | AI 驱动的软件开发平台 | ai, 构建Agent, 拿来就用 | [打开](https://github.com/OpenHands/OpenHands) | | OpenMonoAgent.ai - 开源无限 token 编码 agent,终端原生 | 开源无限 token 编码 agent,终端原生 | agent, open-source, multi-agent, 构建Agent, 拿来就用 | [打开](https://github.com/StartupHakk/OpenMonoAgent.ai) | | OpenSpec - 面向 AI Coding Assistant 的规格驱动开发(SDD)框架 | 面向 AI Coding Assistant 的规格驱动开发(SDD)框架。通过先定义规格、再由 AI 实现代码的方式,提升 AI 编程的可靠性和可预测性。 | framework, coding, 需要开发, 自动化 | [打开](https://github.com/Fission-AI/OpenSpec) | | polpo - 从手机控制Claude Code及其他AI Agent的远程管理工具 | 从手机控制Claude Code及其他AI Agent的远程管理工具 | multi-agent, agent, tool, 拿来就用, 构建Agent | [打开](https://github.com/pugliatechs/polpo) | | portless - 用稳定的命名 URL 替代 localhost 端口号,面向人类和 Agent 的开发工具 | 用稳定的命名 URL 替代 localhost 端口号,面向人类和 Agent 的开发工具。 | tool, multi-agent, agent, 拿来就用, 部署运维 | [打开](https://github.com/vercel-labs/portless) | | proofshot - 为AI编码Agent提供UI验证能力,让Agent能检查自己构建的界面 | 为AI编码Agent提供UI验证能力,让Agent能检查自己构建的界面 | multi-agent, agent, 自动化, 构建Agent | [打开](https://github.com/AmElmo/proofshot) | | public-apis - 公开 API 集合,开发者的必备参考 | 公开 API 集合,开发者的必备参考。 | ai, 学习参考, 需要开发 | [打开](https://github.com/public-apis/public-apis) | | re_gent - 专为 AI Agent 设计的版本控制系统 | 专为 AI Agent 设计的版本控制系统 | multi-agent, agent, 构建Agent, 自动化 | [打开](https://github.com/regent-vcs/re_gent) | | react-doctor - 检测 Agent 生成的 React 代码质量问题,自动诊断和修复建议 | 检测 Agent 生成的 React 代码质量问题,自动诊断和修复建议 | multi-agent, agent, 自动化, 拿来就用 | [打开](https://github.com/millionco/react-doctor) | | resurf - AI 浏览器 Agent 的现实可复现测试框架 | AI 浏览器 Agent 的现实可复现测试框架 | multi-agent, framework, agent, 需要开发, 自动化 | [打开](https://github.com/lightfeed/resurf) | | Scrapling - 自适应网页爬虫框架,支持从单请求到大规模爬取 | 自适应网页爬虫框架,支持从单请求到大规模爬取 | framework, 需要开发, 自动化 | [打开](https://github.com/D4Vinci/Scrapling) | | semble - 为AI Agent设计的高效代码搜索库,静态嵌入实现0.854 NDCG@10准确率,索引250ms查询1.5ms | 为AI Agent设计的高效代码搜索库,静态嵌入实现0.854 NDCG@10准确率,索引250ms查询1.5ms | agent, multi-agent, 需要开发, 构建Agent | [打开](https://github.com/MinishLab/semble) | | shark - 面向 AI Agent 的开源认证系统,Go 语言实现 | 面向 AI Agent 的开源认证系统,Go 语言实现 | multi-agent, open-source, agent, 构建Agent, 需要开发 | [打开](https://github.com/shark-auth/shark) | | spec-kit - GitHub官方Spec-Driven Development工具包 | GitHub官方Spec-Driven Development工具包 | tool, 拿来就用, 自动化 | [打开](https://github.com/github/spec-kit) | | stage-cli - 帮助开发者逐步审查 AI Agent 生成代码变更的 CLI 工具 | 帮助开发者逐步审查 AI Agent 生成代码变更的 CLI 工具 | tool, multi-agent, agent, 拿来就用 | [打开](https://github.com/ReviewStage/stage-cli) | | tessera - 多 Agent 编码工作空间 (Web+桌面) | 多 Agent 编码工作空间 (Web+桌面) | multi-agent, agent, 构建Agent, 拿来就用 | [打开](https://github.com/horang-labs/tessera) | | thr - 本地语义记忆CLI工具,让编码Agent在本地存储和按语义检索记忆,无需服务器和API | 本地语义记忆CLI工具,让编码Agent在本地存储和按语义检索记忆,无需服务器和API | tool, agent, multi-agent, 构建Agent, 拿来就用 | [打开](https://github.com/Chadi00/thr) | | token-optimizer - Token优化工具,发现并修复幽灵token | Token优化工具,发现并修复幽灵token | tool, 拿来就用 | [打开](https://github.com/alexgreensh/token-optimizer) | | traceroot - 分布式服务的开源 Agent 化调试工具 | 分布式服务的开源 Agent 化调试工具 | tool, open-source, agent, multi-agent, 构建Agent, 部署运维 | [打开](https://github.com/traceroot-ai/traceroot) | | TrainForgeTester - AI Agent确定性场景测试工具,检测错误操作、跳过步骤、错误工具调用等问题 | AI Agent确定性场景测试工具,检测错误操作、跳过步骤、错误工具调用等问题 | tool, agent, multi-agent, 构建Agent, 自动化 | [打开](https://github.com/TrainForge/TrainForgeTester) |