{"id":48789175,"url":"https://github.com/ginsing1226/screenclaw","last_synced_at":"2026-04-13T19:00:59.647Z","repository":{"id":347797795,"uuid":"1191584661","full_name":"GinSing1226/ScreenClaw","owner":"GinSing1226","description":"Screenshot + percentage grids enabling any multimodal LLM for non-blocking RPA/Computer Use。为任意多模态大模型提供截图+百分比坐标网格，实现无感无阻塞的RPA和电脑使用","archived":false,"fork":false,"pushed_at":"2026-04-13T17:11:09.000Z","size":28599,"stargazers_count":23,"open_issues_count":0,"forks_count":6,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-04-13T19:00:16.297Z","etag":null,"topics":["agent","claude-code","computer-use","openclaw","python","rpa","skills","vision-language-model"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/GinSing1226.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-03-25T11:45:19.000Z","updated_at":"2026-04-13T17:11:13.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/GinSing1226/ScreenClaw","commit_stats":null,"previous_names":["ginsing1226/screenclaw"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/GinSing1226/ScreenClaw","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/GinSing1226%2FScreenClaw","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/GinSing1226%2FScreenClaw/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/GinSing1226%2FScreenClaw/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/GinSing1226%2FScreenClaw/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/GinSing1226","download_url":"https://codeload.github.com/GinSing1226/ScreenClaw/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/GinSing1226%2FScreenClaw/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":31766482,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-04-13T15:25:13.801Z","status":"ssl_error","status_checked_at":"2026-04-13T15:25:09.162Z","response_time":93,"last_error":"SSL_read: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["agent","claude-code","computer-use","openclaw","python","rpa","skills","vision-language-model"],"created_at":"2026-04-13T19:00:28.371Z","updated_at":"2026-04-13T19:00:59.632Z","avatar_url":"https://github.com/GinSing1226.png","language":"Python","readme":"# ScreenClaw\n\n\u003cdiv align=\"center\"\u003e\n\n**睇虾 — AI 应用的可视化操作必备伴侣**\n\n让任何多模态大模型识别网格坐标，自动化操作任何软件\n\n[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)\n[![Python 3.11+](https://img.shields.io/badge/python-3.11+-blue.svg)](https://www.python.org/downloads/)\n[![Tauri 2.0](https://img.shields.io/badge/Tauri-2.0-ffcf131.svg)](https://tauri.app/)\n\n[English](./README_EN.md) | [Agent Guide](./README_AGENT.md) | [Demo Video](https://www.bilibili.com/video/BV1WJD8BUEnh)\n\n\u003c/div\u003e\n\n---\n\n## 📖 什么是 ScreenClaw？\n\nScreenClaw（睇虾）是一款本地运行的\"中间件\"程序，它充当 AI 应用与桌面软件之间的桥梁。\n\n**核心创新**：通过在截图中叠加百分比坐标网格，让任何多模态大模型都能精确识别界面元素位置，无需专门训练 UI 模型。\n\nAI 应用通过 HTTP API 可以：\n- 📸 **截图并绘制坐标网格** — 精确定位界面元素位置\n- 🖱️ **注入鼠标键盘操作** — 后台无感控制，不影响用户使用\n- 🔄 **自动化任何软件** — 即使软件没有 API 或命令行接口\n\n**关键特性**：全程不抢夺用户的物理鼠标和键盘，你可以在看电影、写文档的同时，让 AI 在后台完成自动化任务。\n\n---\n\n## ✨ 核心特性\n\n### 🖥️ 后台无感操作\n采用 `background` 模式，通过 PostMessage/SendMessage 注入事件，不会激活窗口或抢占焦点，用户毫无感知。\n\n### 📐 百分比坐标网格\n- 让任何多模态大模型有了精确定位能力，且识别速度更快\n- 基于百分比（0-100）的坐标系统，只要窗口比例不变，适用于任何大小的窗口\n- 不同分辨率、不同窗口尺寸都能复用相同的坐标\n\n### 💻 操作电脑软件\n支持自动化操作大部分电脑软件，特别是不提供 API、CLI 等自动化工具的传统桌面软件。\n\n### 📱 操作手机\n支持自动化操作手机模拟器、官方手机助手（各家手机厂商的多屏协同/手机投屏）：\n- 无需 ADB\n- 无需 root\n- 无需虚拟机\n- 无需专门 UI 大模型就能操作手机 APP\n\n### 💎 沉淀复用\n- 支持将成功的任务流程沉淀为场景模板，下次直接复用\n- 接入 OpenClaw 等个人 AI 助手后，个性化数据在本地留存\n- AI 会越来越懂你，效率越来越高\n\n### 🔒 安全可控\n- Token 认证机制\n- 禁止访问进程黑名单，避免 AI 访问敏感应用\n- Hijack 操作需用户确认\n\n### 🎮 托管模式\n用户主动要求时进入，AI 完全控制电脑，物理输入，无需逐次确认。适合需要持续操作的场景（如中文输入）。通过快捷键 `Ctrl+Alt+Z` 可随时退出。\n\n---\n\n## 🚀 快速开始\n\n### 从源码运行（开发环境）\n\n**一行部署和启动**：\n```bash\ngit clone https://github.com/GinSing1226/ScreenClaw.git \u0026\u0026 cd ScreenClaw \u0026\u0026 pip install -r python/requirements.txt \u0026\u0026 npm install \u0026\u0026 npm run tauri dev\n```\n\n**已有项目**：\n```bash\n# 安装依赖（首次或依赖更新时）\npip install -r python/requirements.txt \u0026\u0026 npm install\n\n# 运行开发环境1 建议以管理员身份启动终端\nnpm run tauri dev\n\n# 或运行开发环境2\nnpx tauri dev\n```\n\n### 下载 Release 运行（生产环境）\n\n1. 前往 [Releases](https://github.com/GinSing1226/ScreenClaw/releases) 页面\n2. 下载最新版本的 Windows 压缩包\n3. 解压后双击 `ScreenClaw.exe` 启动\n\n### 启动服务\n\n1. 应用启动后，点击\"启动服务\"按钮\n2. 查看连接信息：\n   - 本机访问：`http://127.0.0.1:12261`\n   - 局域网访问：`http://192.168.x.x:12261`\n\n### 安装 Skill\n\n在 Claude Code、OpenClaw、OpenCode、Codex 等 AI Agent 工具中执行：\n\n```bash\nnpx skills add GinSing1226/ScreenClaw\n```\n\n安装后，AI 即可自动调用 ScreenClaw API 进行桌面软件自动化操作。\n\n---\n\n## 📚 API 清单\n\n| API | 用途 | 说明 |\n|-----|------|------|\n| `health` | 验证服务是否可连接 | 操作前的第一步检查 |\n| `get_window_list` | 查找目标窗口 | 获取 window_id（后续操作必需） |\n| `screenshot` | 查看界面状态、定位坐标 | 每次操作后验证结果 |\n| `click` | 触发功能、进入页面 | 普通点击操作 |\n| `long_press` | 触发长按功能 | 拖拽起点、显示菜单等 |\n| `swipe` | 触摸式滑动 | 翻页、切换标签、拖拽 |\n| `scroll` | 鼠标滚轮滚动 | 浏览长内容、列表 |\n| `right_click` | 打开上下文菜单 | 调用快捷方式 |\n| `hover` | 触发悬停效果 | 显示 tooltip、隐藏 UI 元素 |\n| `input_text` | 输入文本内容 | 填表、搜索等 |\n| `press_key` | 触发快捷键 | Ctrl+C 复制、Enter 确认等 |\n| `wait` | 等待 UI 稳定 | 等待动画、加载完成 |\n| `batch` | 执行多步骤流程 | 减少网络请求、固定序列 |\n| `drag` | 拖拽操作 | 文件拖放、元素拖动，支持速度控制 |\n| `mouse_move` | 鼠标相对移动 | 游戏视角控制，仅 hijack/delegated |\n| `scroll_screenshot` | 滚动长截图 | 自动滚动拼接，生成长图 |\n| `delegated` | 进入/退出托管模式 | 用户主动要求完全控制时 |\n\n---\n\n## 🔧 API 使用示例\n\n### 通用请求头\n\n所有 API 请求都需要携带：\n\n```bash\nAuthorization: Bearer {token}\nContent-Type: application/json\n```\n\n### 1. 健康检查\n\n```bash\ncurl -X GET http://127.0.0.1:12261/api/health\n```\n\n### 2. 获取窗口列表\n\n```bash\ncurl -X POST http://127.0.0.1:12261/api/get_window_list \\\n  -H \"Authorization: Bearer {token}\" \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\n    \"ai_app_type\": \"claude_code\",\n    \"session_id\": \"session-123\",\n    \"keyword\": \"notepad\",\n    \"include_children\": true,\n    \"children_filter\": \"titled\"\n  }'\n```\n\n### 3. 截图（带坐标网格）\n\n```bash\ncurl -X POST http://127.0.0.1:12261/api/screenshot \\\n  -H \"Authorization: Bearer {token}\" \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\n    \"ai_app_type\": \"claude_code\",\n    \"session_id\": \"session-123\",\n    \"window_id\": 1001,\n    \"main_window_id\": 1001,\n    \"coordinate_type\": \"grid\",\n    \"color_mode\": \"grayscale\",\n    \"density\": 5.0,\n    \"opacity\": 50,\n    \"color\": \"#ff0000\",\n    \"number_density\": 2,\n    \"number_decimal\": 0,\n    \"number_size\": 12,\n    \"number_color\": \"#ff0000\",\n    \"number_opacity\": 100\n  }'\n```\n\n### 4. 点击\n\n```bash\ncurl -X POST http://127.0.0.1:12261/api/click \\\n  -H \"Authorization: Bearer {token}\" \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\n    \"ai_app_type\": \"claude_code\",\n    \"session_id\": \"session-123\",\n    \"window_id\": 1001,\n    \"main_window_id\": 1001,\n    \"x\": 50.0,\n    \"y\": 30.0,\n    \"action_method\": \"background\"\n  }'\n```\n\n### 5. 长按\n\n```bash\ncurl -X POST http://127.0.0.1:12261/api/long_press \\\n  -H \"Authorization: Bearer {token}\" \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\n    \"ai_app_type\": \"claude_code\",\n    \"session_id\": \"session-123\",\n    \"window_id\": 1001,\n    \"main_window_id\": 1001,\n    \"x\": 50.0,\n    \"y\": 50.0,\n    \"duration_ms\": 500,\n    \"action_method\": \"background\"\n  }'\n```\n\n### 6. 滑动\n\n```bash\ncurl -X POST http://127.0.0.1:12261/api/swipe \\\n  -H \"Authorization: Bearer {token}\" \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\n    \"ai_app_type\": \"claude_code\",\n    \"session_id\": \"session-123\",\n    \"window_id\": 1001,\n    \"main_window_id\": 1001,\n    \"start_x\": 50.0,\n    \"start_y\": 80.0,\n    \"end_x\": 50.0,\n    \"end_y\": 20.0,\n    \"action_method\": \"background\"\n  }'\n```\n\n### 7. 滚动\n\n```bash\ncurl -X POST http://127.0.0.1:12261/api/scroll \\\n  -H \"Authorization: Bearer {token}\" \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\n    \"ai_app_type\": \"claude_code\",\n    \"session_id\": \"session-123\",\n    \"window_id\": 1001,\n    \"main_window_id\": 1001,\n    \"x\": 50.0,\n    \"y\": 50.0,\n    \"delta\": -120,\n    \"action_method\": \"background\"\n  }'\n```\n\n### 8. 右键点击\n\n```bash\ncurl -X POST http://127.0.0.1:12261/api/right_click \\\n  -H \"Authorization: Bearer {token}\" \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\n    \"ai_app_type\": \"claude_code\",\n    \"session_id\": \"session-123\",\n    \"window_id\": 1001,\n    \"main_window_id\": 1001,\n    \"x\": 50.0,\n    \"y\": 50.0,\n    \"action_method\": \"background\"\n  }'\n```\n\n### 9. 鼠标悬浮\n\n```bash\ncurl -X POST http://127.0.0.1:12261/api/hover \\\n  -H \"Authorization: Bearer {token}\" \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\n    \"ai_app_type\": \"claude_code\",\n    \"session_id\": \"session-123\",\n    \"window_id\": 1001,\n    \"main_window_id\": 1001,\n    \"x\": 50.0,\n    \"y\": 50.0,\n    \"duration_ms\": 500,\n    \"action_method\": \"hijack\"\n  }'\n```\n\n### 10. 输入文本\n\n```bash\ncurl -X POST http://127.0.0.1:12261/api/input_text \\\n  -H \"Authorization: Bearer {token}\" \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\n    \"ai_app_type\": \"claude_code\",\n    \"session_id\": \"session-123\",\n    \"window_id\": 1001,\n    \"main_window_id\": 1001,\n    \"x\": 50.0,\n    \"y\": 50.0,\n    \"text\": \"Hello World\\n\",\n    \"action_method\": \"hijack\"\n  }'\n```\n\n### 11. 按键\n\n```bash\ncurl -X POST http://127.0.0.1:12261/api/press_key \\\n  -H \"Authorization: Bearer {token}\" \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\n    \"ai_app_type\": \"claude_code\",\n    \"session_id\": \"session-123\",\n    \"window_id\": 1001,\n    \"main_window_id\": 1001,\n    \"key\": \"ctrl c\",\n    \"action_method\": \"background\"\n  }'\n```\n\n### 12. 等待\n\n```bash\ncurl -X POST http://127.0.0.1:12261/api/wait \\\n  -H \"Authorization: Bearer {token}\" \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\n    \"ai_app_type\": \"claude_code\",\n    \"session_id\": \"session-123\",\n    \"window_id\": 1001,\n    \"duration_ms\": 1000\n  }'\n```\n\n### 13. 批量执行\n\n```bash\ncurl -X POST http://127.0.0.1:12261/api/batch \\\n  -H \"Authorization: Bearer {token}\" \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\n    \"ai_app_type\": \"claude_code\",\n    \"session_id\": \"session-123\",\n    \"window_id\": 1001,\n    \"main_window_id\": 1001,\n    \"instructions\": [\n      {\"action\": \"click\", \"params\": {\"x\": 50, \"y\": 35, \"action_method\": \"background\"}},\n      {\"action\": \"wait\", \"params\": {\"duration_ms\": 200}},\n      {\"action\": \"input_text\", \"params\": {\"x\": 50, \"y\": 35, \"text\": \"hello\", \"action_method\": \"hijack\"}},\n      {\"action\": \"wait\", \"params\": {\"duration_ms\": 300}},\n      {\"action\": \"screenshot\", \"params\": {\"coordinate_type\": \"no\"}}\n    ]\n  }'\n```\n\n### 14. 托管模式\n\n```bash\ncurl -X POST http://127.0.0.1:12261/api/delegated \\\n  -H \"Authorization: Bearer {token}\" \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\n    \"ai_app_type\": \"claude_code\",\n    \"session_id\": \"session-123\",\n    \"action\": \"enter\"\n  }'\n```\n\n退出托管模式：将 `\"action\": \"enter\"` 改为 `\"action\": \"exit\"`，或按快捷键 `Ctrl+Alt+Z`。\n\n### 15. 拖拽\n\n```bash\ncurl -X POST http://127.0.0.1:12261/api/drag \\\n  -H \"Authorization: Bearer {token}\" \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\n    \"ai_app_type\": \"claude_code\",\n    \"session_id\": \"session-123\",\n    \"window_id\": 1001,\n    \"main_window_id\": 1001,\n    \"start_x\": 30.0,\n    \"start_y\": 50.0,\n    \"end_x\": 70.0,\n    \"end_y\": 50.0,\n    \"duration_ms\": 500,\n    \"action_method\": \"background\"\n  }'\n```\n\n### 16. 鼠标移动\n\n```bash\ncurl -X POST http://127.0.0.1:12261/api/mouse_move \\\n  -H \"Authorization: Bearer {token}\" \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\n    \"ai_app_type\": \"claude_code\",\n    \"session_id\": \"session-123\",\n    \"window_id\": 1001,\n    \"main_window_id\": 1001,\n    \"delta_x\": 200,\n    \"delta_y\": 0,\n    \"duration_ms\": 300,\n    \"action_method\": \"hijack\"\n  }'\n```\n\n### 17. 滚动长截图\n\n```bash\ncurl -X POST http://127.0.0.1:12261/api/scroll_screenshot \\\n  -H \"Authorization: Bearer {token}\" \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\n    \"ai_app_type\": \"claude_code\",\n    \"session_id\": \"session-123\",\n    \"window_id\": 1001,\n    \"main_window_id\": 1001,\n    \"max_scrolls\": 20,\n    \"scroll_percent\": 0.85,\n    \"scroll_wait\": 1.0,\n    \"action_method\": \"hijack\"\n  }'\n```\n\n---\n\n## 🎯 适用场景\n\n### 自动化可视化测试\n模拟人工操作可视化界面，沉淀测试用例，自动打开 F12 管理报错。\n\n### 自动操作无网页版、无 API、无 CLI 的电脑软件\n传统 ERP、OA 系统、专业工具软件等，只要能显示在屏幕上就能操作。\n\n### 手机模拟器、手机投屏助手\n对于大部分普通人，现阶段的 ADB、虚拟机、docker 等方案的自动化手机太麻烦，门槛太高。专门的 AI 手机又已停产。ScreenClaw 让你直接操作投屏/模拟器界面。\n\n### 复杂任务自动化\n专用的 UI 大模型通常参数量较低，难以理解复杂任务，难以尝试长时程执行任务。而 ScreenClaw 能让 SOTA 级别的多模态大模型（如 GPT-4V、Claude 3.5 Sonnet）自动化操作软件。\n\n### 与其他工具对比\n\n| 场景 | 推荐工具 | ScreenClaw 作用 |\n|------|----------|----------------|\n| 浏览器自动化 | Playwright、CDP、agent-browser | 补充：模拟人工走查操作、获取控制台报错 |\n| 传统 RPA | UiPath、Power Automate | 更智能：AI 自动识别坐标，无需人工配置 |\n\n---\n\n## 💡 平台说明\n\n### 执行方（运行 ScreenClaw 的机器）\n**推荐：Windows**\n\n- ✅ 支持无感截图（后台截图，不抢占前台）\n- ✅ 支持无感鼠标和键盘操作（不影响用户使用）\n- ✅ 拥有丰富的手机模拟器（如 MuMu、蓝叠、夜神等）\n- ✅ 各大手机厂商的投屏软件（华为多屏协同、小米跨屏协作等）\n- ✅ 能自动化大部分手机 APP\n\n### 调用方（运行 AI 应用的机器）\n**支持：Windows / macOS / Linux**\n\n通过 HTTP API 调用，跨平台无限制。\n\n### 部署模式\n- **本地模式**：执行方和调用方是同一台机器\n- **局域网模式**：不同机器，需在同一局域网内\n\n---\n\n## 🏗️ 技术架构\n\n```\n┌─────────────────────────────────────────────┐\n│         外部 AI 应用                         │\n│  (Claude Code / openclaw / 其他 Agent)         │\n└──────────────────┬──────────────────────────┘\n                   │ HTTP API\n                   ▼\n┌─────────────────────────────────────────────┐\n│         Python 后端服务                      │\n│       FastAPI :12261                         │\n│  ┌───────┬────────┬────────┬────────┐      │\n│  │ API层 │ Core层 │Platform│Service │      │\n│  └───────┴────────┴────────┴────────┘      │\n└──────────────────┬──────────────────────────┘\n                   │ 子进程\n┌──────────────────▼──────────────────────────┐\n│         Tauri 桌面应用                       │\n│  监控面板 │ 系统设置 │ 托盘图标               │\n└─────────────────────────────────────────────┘\n```\n\n### 技术栈\n\n| 层级 | 技术 |\n|------|------|\n| 前端 | Tauri 2.0 + Vue 3 + TypeScript |\n| 后端 | Python 3.11 + FastAPI + Uvicorn |\n| 截图 | pywin32 / mss |\n| 操作注入 | pywin32 (PostMessage) / pyautogui (SendInput) |\n| 网格绘制 | Pillow (PIL) |\n\n---\n\n## 📁 项目结构\n\n```\nscreenClaw/\n├── python/                   # Python 后端\n│   ├── app/\n│   │   ├── api/             # API 路由\n│   │   ├── core/            # 核心业务逻辑\n│   │   ├── platform/        # 平台适配层\n│   │   ├── models/          # 数据模型\n│   │   ├── services/        # 服务层\n│   │   ├── scripts/         # API 调用脚本\n│   │   └── utils/           # 工具函数\n│   └── main.py              # 入口文件\n│\n├── src-tauri/               # Tauri 后端（Rust）\n│   └── src/\n│       ├── commands.rs      # Tauri 命令\n│       └── main.rs          # 入口\n│\n├── src/                     # Vue 前端\n│   ├── components/          # 组件\n│   ├── composables/         # 组合式函数\n│   └── main.ts              # 入口\n│\n├── skills/                  # AI Skill\n│   └── screenclaw/\n│       ├── SKILL.md         # 技能定义（完整使用指南）\n│       └── references/      # API 文档和场景模板\n│\n└── data/                    # 数据目录\n    └── config.json          # 配置文件（自动创建）\n```\n\n---\n\n## ⚙️ 配置说明\n\n### ScreenClaw 服务配置\n位于 `data/config.json`（应用初始化时自动创建）：\n\n```json\n{\n  \"server\": {\n    \"host\": \"0.0.0.0\",\n    \"port\": 12261,\n    \"token\": \"your-token-here\",\n    \"local_ip\": \"192.168.x.x\"\n  }\n}\n```\n\n### AI Skill 连接配置\n位于技能目录 `{skill-dir}/references/config.md`，用于 AI 应用连接 ScreenClaw 服务。\n\n```json\n{\n  \"server\": {\n    \"host\": \"0.0.0.0\",\n    \"port\": 12261,\n    \"token\": \"your-token-here\",\n    \"local_ip\": \"192.168.x.x\"\n  }\n}\n```\n\n| 配置项 | 说明 |\n|--------|------|\n| `host` | 绑定地址，0.0.0.0 允许局域网访问 |\n| `port` | HTTP 服务端口 |\n| `token` | API 认证令牌 |\n| `local_ip` | 自动检测的局域网 IP |\n\n---\n\n## 🔒 安全说明\n\n- 所有 API 请求需要 Token 认证\n- Hijack 模式操作会弹出确认窗口\n- 可配置禁止访问的进程黑名单\n- 仅允许局域网和本机访问\n\n---\n\n## 📖 更多文档\n\n- [Skill 使用指南](./skills/screenclaw/SKILL.md) — 完整的 API 文档和使用方法论\n\n---\n\n## 🗺️ 未来规划\n\n- 🎯 **提高识别率** — 通过技能优化和场景模板沉淀，提升多模态大模型的初次识别准确率\n- 🔄 **更多 RPA 动作** — 扩展支持的自动化操作类型，覆盖更多使用场景\n- 🖥️ **桌面级操作** — 直接对桌面截图和操作，不绑定进程，可操作任何可见内容（会抢夺用户鼠标和键盘）\n- 📦 **场景模板库** — 封装更多常用软件的场景模板，开箱即用\n\n---\n\n## 🤝 贡献指南\n\n欢迎提交 Issue 和 Pull Request！\n\n---\n\n## 📄 许可证\n\nMIT License\n\n---\n\n\u003cdiv align=\"center\"\u003e\n\n**如果这个项目对你有帮助，请给一个 ⭐ Star 支持一下！**\n\n**睇虾 — 让任何多模态大模型操作任何屏幕**\n\nMade with ❤️ by GinSing\n\n\u003c/div\u003e\n","funding_links":[],"categories":[],"sub_categories":[],"project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fginsing1226%2Fscreenclaw","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fginsing1226%2Fscreenclaw","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fginsing1226%2Fscreenclaw/lists"}