webcode语言:中文 | English webcode 用来把 ChatGPT、Gemini、DeepSeek 等这类网页 AI 接到本地 VS Code 中,支持读、写、编辑本地文件,运行命令,还支持 MCP 和 Skills。 快速开始1. 安装 VS Code 扩展
2. 启动 Gateway
当状态栏显示端口号时,说明本地 Gateway 已经启动。启动成功后会自动打开 AI 启动菜单,不需要再点一次状态栏。 3. 打开目标网页 AI
首次使用时,需要在这个独立 Edge profile 中登录一次目标 AI 站点。登录过程中如果跳转到 Google、Microsoft 等第三方登录页,bridge 会暂时暂停当前页面能力并保留 session;登录完成并回到目标 AI 站点后会自动恢复。当 bridge 显示 如果 Gateway 已经在运行,点击状态栏里的 4. 在对话中使用
如果初始化上下文超过当前 AI 站点的输入框限制,webcode 会优先把完整上下文作为 txt 附件添加到消息中,并在输入框里保留一段简短说明。 例如:
内置工具webcode 内置 具体工具清单见 内置工具说明。 MCP需要接入浏览器自动化、GitHub、数据库或其他外部能力时,可以通过 具体配置方式见 MCP 服务配置指南。 Skillswebcode 可以把当前 VS Code 工作区中的本地 Skills 暴露给网页 AI,让模型按需读取项目工作流、模板、领域说明或脚本资源。 具体使用方式见 Skills 使用指南。 项目规则初始化时,webcode 会从 VS Code 当前打开的主工作区根目录读取项目规则文件,并随初始化结果发送给网页 AI:
如果 其他 AI 平台支持webcode 内置支持常见网页 AI,也允许通过 VS Code 配置快速接入新的站点。稳定平台可以加入内置站点列表,实验或私有站点通常只需要配置 具体步骤见 站点支持扩展指南。 从源码构建源码下载、依赖安装、打包脚本、产物说明和本地安装方式见 构建指南。 常见问题触发词无反应、工具调用执行后页面没有变化、查看历史工具调用等问题见 常见问题。 参与贡献欢迎提交 Issue 和 Pull Request。 许可证webcodeLanguage: English | 中文 webcode connects ChatGPT, Gemini, DeepSeek, and other web AI products to local VS Code so they can read, write, and edit local files, run commands, and use MCP and Skills. Quick Start1. Install the VS Code Extension
2. Start the Gateway
When the status bar shows a port number, the local gateway is running. The AI launch menu opens automatically after startup, so you do not need to click the status bar a second time. 3. Open a Supported AI Product
On first use, sign in to the target AI site once in the isolated Edge profile. If sign-in redirects to a third-party provider such as Google or Microsoft, the bridge temporarily pauses page capabilities and keeps the session; it resumes automatically after the browser returns to the target AI site. When the bridge extension shows If the gateway is already running, click 4. Use It in Chat
If the initialization context exceeds the current AI site's input limit, webcode first attaches the full context as a txt file and keeps a short instruction in the input box. For example:
Built-in Toolswebcode includes built-in local tools such as See the built-in tools reference. MCPUse See the MCP server configuration guide. Skillswebcode can expose local Skills from the current VS Code workspace so the web AI can load project workflows, templates, domain notes, or script resources on demand. See the Skills guide. Project RulesDuring initialization, webcode reads project rule files from the root of the primary VS Code workspace and sends them with the initialization result:
If both Additional AI Platform Supportwebcode includes built-in support for common web AI products and can also attach new sites through VS Code configuration. Stable platforms can be added as built-in sites; experimental or private sites usually only need See the site support guide. Build From SourceSee the build guide for repository checkout, dependency installation, packaging scripts, artifacts, and local installation. FAQSee the FAQ guide for trigger-word issues, tool calls that executed but did not update the page, and tool-call history logs. ContributingIssues and pull requests are welcome. License |