[{"data":1,"prerenderedAt":1738},["ShallowReactive",2],{"header-counts":3,"compare-autoglm-vs-manus":6,"footer-counts":7,"compare-a-autoglm":10,"compare-b-manus":747},{"tools":4,"reviews":5},77,25,null,{"tools":4,"reviews":5,"playbooks":8,"news":9},22,13,{"id":11,"title":12,"alternatives":13,"api_compatible":6,"body":17,"category":676,"chinese_friendly":264,"cover":677,"description":678,"domestic":679,"extension":680,"faq":681,"free":679,"github":6,"languages":694,"meta":697,"models":6,"navigation":253,"notSuitable":6,"opensource":253,"path":698,"pillar":699,"platforms":700,"priceTable":704,"pricing":718,"published":719,"relatedPlaybooks":720,"relatedReviews":6,"score":722,"self_host":253,"seo":723,"seoTitle":6,"slug":724,"sources":725,"stem":735,"suitable":6,"tagline":736,"tags":737,"updated":728,"verdict":745,"website":636,"__hash__":746},"tools\u002Ftools\u002Fagent\u002Fdesktop\u002Fautoglm.md","AutoGLM",[14,15,16],"agent\u002Fdesktop\u002Fclaude-desktop","agent\u002Fdesktop\u002Fopenclaw","coding\u002Fagent\u002Fmanus",{"type":18,"value":19,"toc":664},"minimark",[20,25,29,32,35,106,109,129,135,139,144,164,169,192,196,361,364,518,521,571,575,601,604,625,628,660],[21,22,24],"h2",{"id":23},"tldr","TL;DR",[26,27,28],"p",{},"AutoGLM 是智谱清言 + 清华大学合作的 GUI 自治智能体研究项目（arXiv 2411.00820）。2025-12-08 开源 Open-AutoGLM：AutoGLM-Phone-9B 模型（MIT 协议，HuggingFace \u002F ModelScope 可下载）+ Apache 2.0 框架代码 + Android 适配 + Remote ADB + 50+ 中文 App 示例。VAB-WebArena-Lite 55.2% \u002F 59.1% 成功率，覆盖微信 \u002F 淘宝 \u002F 美团 \u002F Gmail \u002F Google Maps 等。Web 端通过智谱清言 Chrome 扩展（10 万+ 用户）提供页面总结 \u002F 高级检索能力。",[26,30,31],{},"适合：研究 GUI agent 的学术 \u002F 工程团队；国内 App 自治场景；要中文模型 + 私有化部署；想本地跑 phone agent 不依赖外网。不适合：跨平台桌面自动化（用 Anthropic Computer Use \u002F Claude Desktop）；纯英文 web 任务（成功率不如商业产品）；要 production-grade 稳定性（学术 \u002F 开源版本仍在演进）。",[21,33,34],{"id":34},"核心能力",[36,37,38,46,52,58,64,70,76,82,88,94,100],"ul",{},[39,40,41,45],"li",{},[42,43,44],"strong",{},"AutoGLM-Phone-9B","：MIT 协议开源模型，HuggingFace \u002F ModelScope 可下载",[39,47,48,51],{},[42,49,50],{},"Android 适配层","：原生 Android API + Accessibility Service",[39,53,54,57],{},[42,55,56],{},"Remote ADB","：初次配置后通过 WiFi 控制设备，无需 USB",[39,59,60,63],{},[42,61,62],{},"多模态屏幕理解","：截屏 + UI 树双输入，规划 + grounding 分离",[39,65,66,69],{},[42,67,68],{},"50+ 中文 App 任务示例","：微信 \u002F 淘宝 \u002F 美团 \u002F 京东 \u002F 支付宝 \u002F 抖音 \u002F 小红书 \u002F 大众点评等",[39,71,72,75],{},[42,73,74],{},"英文 App 覆盖","：Gmail \u002F Google Maps \u002F X \u002F Reddit \u002F OneStopShop",[39,77,78,81],{},[42,79,80],{},"Web 浏览器智能体","：Chrome 扩展 + VAB-WebArena-Lite 55.2% \u002F 59.1%",[39,83,84,87],{},[42,85,86],{},"页面总结 + 划线助手 + 写作助手 + 高级检索","（知网 \u002F 知乎 \u002F 小红书）",[39,89,90,93],{},[42,91,92],{},"Safe Operations","：敏感操作（登录 \u002F 验证码）确认后由人接管",[39,95,96,99],{},[42,97,98],{},"Python API","：几行代码就能跑通自动化任务",[39,101,102,105],{},[42,103,104],{},"私有化部署","：模型 + 框架本地跑，数据 \u002F 日志 \u002F 权限完全自控",[21,107,108],{"id":108},"价格",[36,110,111,117,123],{},[39,112,113,116],{},[42,114,115],{},"Open-AutoGLM","：$0；模型 MIT + 代码 Apache 2.0",[39,118,119,122],{},[42,120,121],{},"Chrome 扩展","：$0；Chrome Web Store 装即用",[39,124,125,128],{},[42,126,127],{},"整合产品（智谱清言 App）","：商业服务，价格随订阅",[130,131,132],"blockquote",{},[26,133,134],{},"真实成本 = 0（开源）+ GPU 推理（9B 模型 RTX 4090 一张即可跑）",[21,136,138],{"id":137},"实测开源版本-pixel-6a-ubuntu-2204","实测（开源版本 \u002F Pixel 6a + Ubuntu 22.04）",[26,140,141],{},[42,142,143],{},"亮点：",[36,145,146,149,152,155,158,161],{},[39,147,148],{},"32 个月研发的『手机 GUI agent』终于开源，是国内学术圈难得的开放姿态",[39,150,151],{},"中文 App 覆盖度领先：微信 \u002F 美团 \u002F 淘宝跑通率高",[39,153,154],{},"模型 9B 可在单卡 4090 + 24GB 显存跑，门槛友好",[39,156,157],{},"Remote ADB 让长跑实验摆脱 USB 线",[39,159,160],{},"学术 benchmark 数据透明（55.2% \u002F 59.1%）",[39,162,163],{},"safe-operation 设计对登录 \u002F 支付环节做人审接管，避免误操作",[26,165,166],{},[42,167,168],{},"踩坑：",[36,170,171,174,177,180,183,186,189],{},[39,172,173],{},"App 版本更新后 UI 元素变化要重新适配",[39,175,176],{},"9B 模型在复杂多步骤任务上推理不如商业大模型（GPT-5.4 \u002F Claude Opus 4.6）",[39,178,179],{},"主要面向 Android，iOS 支持空缺",[39,181,182],{},"文档以 README + 论文为主，工程化最佳实践薄",[39,184,185],{},"国行 App 隐私 \u002F 反爬监测会偶尔拦截",[39,187,188],{},"Chrome 扩展 3.4 \u002F 5 评分中等，功能尚可但稳定性一般",[39,190,191],{},"整合版（智谱清言）和开源版差异要注意，别混用",[21,193,195],{"id":194},"上手open-autoglm","上手（Open-AutoGLM）",[197,198,203],"pre",{"className":199,"code":200,"language":201,"meta":202,"style":202},"language-bash shiki shiki-themes github-light github-dark","git clone https:\u002F\u002Fgithub.com\u002Fzai-org\u002FOpen-AutoGLM\ncd Open-AutoGLM && pip install -r requirements.txt\n\n# 下载 9B 模型\nhuggingface-cli download THUDM\u002FAutoGLM-Phone-9B\n\n# 配置 ADB（USB 一次）\nadb tcpip 5555 && adb connect \u003Cdevice-ip>:5555\n\n# 跑示例任务\npython examples\u002Frun_phone_agent.py \\\n  --model THUDM\u002FAutoGLM-Phone-9B \\\n  --task \"在小红书搜索罗马旅游攻略并总结景点\"\n","bash","",[204,205,206,222,248,255,262,274,279,285,319,324,330,342,353],"code",{"__ignoreMap":202},[207,208,211,215,219],"span",{"class":209,"line":210},"line",1,[207,212,214],{"class":213},"sScJk","git",[207,216,218],{"class":217},"sZZnC"," clone",[207,220,221],{"class":217}," https:\u002F\u002Fgithub.com\u002Fzai-org\u002FOpen-AutoGLM\n",[207,223,225,229,232,236,239,242,245],{"class":209,"line":224},2,[207,226,228],{"class":227},"sj4cs","cd",[207,230,231],{"class":217}," Open-AutoGLM",[207,233,235],{"class":234},"sVt8B"," && ",[207,237,238],{"class":213},"pip",[207,240,241],{"class":217}," install",[207,243,244],{"class":227}," -r",[207,246,247],{"class":217}," requirements.txt\n",[207,249,251],{"class":209,"line":250},3,[207,252,254],{"emptyLinePlaceholder":253},true,"\n",[207,256,258],{"class":209,"line":257},4,[207,259,261],{"class":260},"sJ8bj","# 下载 9B 模型\n",[207,263,265,268,271],{"class":209,"line":264},5,[207,266,267],{"class":213},"huggingface-cli",[207,269,270],{"class":217}," download",[207,272,273],{"class":217}," THUDM\u002FAutoGLM-Phone-9B\n",[207,275,277],{"class":209,"line":276},6,[207,278,254],{"emptyLinePlaceholder":253},[207,280,282],{"class":209,"line":281},7,[207,283,284],{"class":260},"# 配置 ADB（USB 一次）\n",[207,286,288,291,294,297,299,301,304,308,311,313,316],{"class":209,"line":287},8,[207,289,290],{"class":213},"adb",[207,292,293],{"class":217}," tcpip",[207,295,296],{"class":227}," 5555",[207,298,235],{"class":234},[207,300,290],{"class":213},[207,302,303],{"class":217}," connect",[207,305,307],{"class":306},"szBVR"," \u003C",[207,309,310],{"class":217},"device-i",[207,312,26],{"class":234},[207,314,315],{"class":306},">",[207,317,318],{"class":217},":5555\n",[207,320,322],{"class":209,"line":321},9,[207,323,254],{"emptyLinePlaceholder":253},[207,325,327],{"class":209,"line":326},10,[207,328,329],{"class":260},"# 跑示例任务\n",[207,331,333,336,339],{"class":209,"line":332},11,[207,334,335],{"class":213},"python",[207,337,338],{"class":217}," examples\u002Frun_phone_agent.py",[207,340,341],{"class":227}," \\\n",[207,343,345,348,351],{"class":209,"line":344},12,[207,346,347],{"class":227},"  --model",[207,349,350],{"class":217}," THUDM\u002FAutoGLM-Phone-9B",[207,352,341],{"class":227},[207,354,355,358],{"class":209,"line":9},[207,356,357],{"class":227},"  --task",[207,359,360],{"class":217}," \"在小红书搜索罗马旅游攻略并总结景点\"\n",[21,362,363],{"id":363},"对比",[365,366,367,388],"table",{},[368,369,370],"thead",{},[371,372,373,377,379,382,385],"tr",{},[374,375,376],"th",{},"维度",[374,378,12],{},[374,380,381],{},"Claude Desktop",[374,383,384],{},"OpenClaw",[374,386,387],{},"Anthropic Computer Use",[389,390,391,409,426,443,457,473,487,501],"tbody",{},[371,392,393,397,400,403,406],{},[394,395,396],"td",{},"平台",[394,398,399],{},"Android + Web",[394,401,402],{},"macOS \u002F Windows + MCP",[394,404,405],{},"macOS\u002FLinux\u002FWin",[394,407,408],{},"跨平台",[371,410,411,414,417,420,423],{},[394,412,413],{},"开源",[394,415,416],{},"✅ MIT 模型 + Apache 代码",[394,418,419],{},"闭源",[394,421,422],{},"✅ MIT",[394,424,425],{},"❌",[371,427,428,431,434,437,440],{},[394,429,430],{},"中文 App",[394,432,433],{},"✅ 50+ 示例",[394,435,436],{},"–",[394,438,439],{},"部分",[394,441,442],{},"弱",[371,444,445,448,451,453,455],{},[394,446,447],{},"Phone agent",[394,449,450],{},"✅ 9B 模型",[394,452,425],{},[394,454,425],{},[394,456,425],{},[371,458,459,462,464,467,470],{},[394,460,461],{},"桌面 agent",[394,463,425],{},[394,465,466],{},"✅ MCP",[394,468,469],{},"✅ Gateway",[394,471,472],{},"✅ 原生",[371,474,475,478,481,483,485],{},[394,476,477],{},"学术 benchmark",[394,479,480],{},"✅ 公开",[394,482,436],{},[394,484,436],{},[394,486,439],{},[371,488,489,492,494,496,499],{},[394,490,491],{},"整合 IM",[394,493,436],{},[394,495,436],{},[394,497,498],{},"✅ 20+ 平台",[394,500,436],{},[371,502,503,506,509,512,515],{},[394,504,505],{},"适合",[394,507,508],{},"国内 App \u002F 研究 \u002F 私有化",[394,510,511],{},"MCP 生态桌面",[394,513,514],{},"IM 触发个人 agent",[394,516,517],{},"跨平台桌面",[21,519,520],{"id":520},"避坑",[36,522,523,529,535,541,547,553,559,565],{},[39,524,525,528],{},[42,526,527],{},"App 版本敏感","：UI 改动后效果下降，要 pin App 版本或重训样本",[39,530,531,534],{},[42,532,533],{},"9B 模型推理上限","：复杂多步任务上不及 GPT-5.4 \u002F Claude，可叠加外部规划模型",[39,536,537,540],{},[42,538,539],{},"iOS 暂缺","：需要 iOS agent 走别的方案",[39,542,543,546],{},[42,544,545],{},"Remote ADB 安全","：开放端口必须限内网 + 防火墙",[39,548,549,552],{},[42,550,551],{},"国行 App 反爬","：账号风控 \u002F 验证码会触发，必要时 human-in-the-loop",[39,554,555,558],{},[42,556,557],{},"学术 vs 整合版","：开源版能力 ≠ 智谱清言 App 内置版能力，PoC 别混淆",[39,560,561,564],{},[42,562,563],{},"隐私 \u002F 合规","：本地部署是优势，但读取 IM \u002F 通讯录仍要符合用户授权 + 当地法律",[39,566,567,570],{},[42,568,569],{},"MIT 模型 + Apache 代码","：商用务必看 LICENSE，企业合规审计要走流程",[21,572,574],{"id":573},"适合-不适合","适合 \u002F 不适合",[36,576,577,580,583,586,589,592,595,598],{},[39,578,579],{},"✅ GUI agent 学术研究 \u002F 工程团队",[39,581,582],{},"✅ 国内 App 自治（微信 \u002F 美团 \u002F 淘宝 \u002F 小红书）",[39,584,585],{},"✅ 想本地跑 phone agent + 数据自主",[39,587,588],{},"✅ 私有化部署 + 合规需求",[39,590,591],{},"❌ 跨平台桌面自动化（Claude Desktop \u002F Anthropic Computer Use）",[39,593,594],{},"❌ iOS App 自治",[39,596,597],{},"❌ 要 production-grade 稳定性（仍在演进）",[39,599,600],{},"❌ 不懂 ADB \u002F Python 部署的纯用户",[21,602,603],{"id":603},"相关阅读",[36,605,606,613,619],{},[39,607,608],{},[609,610,612],"a",{"href":611},"\u002Ftools\u002Fagent\u002Fdesktop\u002Fclaude-desktop","Claude Desktop 评测",[39,614,615],{},[609,616,618],{"href":617},"\u002Ftools\u002Fagent\u002Fdesktop\u002Fopenclaw","OpenClaw 评测",[39,620,621],{},[609,622,624],{"href":623},"\u002Ftools\u002Fcoding\u002Fagent\u002Fmanus","Manus 评测",[21,626,627],{"id":627},"来源",[629,630,631,639,646,653],"ol",{},[39,632,633,634],{},"AutoGLM 项目主页（含 Open-AutoGLM 2025-12-08 发布说明 \u002F benchmark 数据）",[609,635,636],{"href":636,"rel":637},"https:\u002F\u002Fxiao9905.github.io\u002FAutoGLM",[638],"nofollow",[39,640,641,642],{},"腾讯新闻 — 智谱开源 AutoGLM（32 个月研发 \u002F MIT + Apache 2.0）",[609,643,644],{"href":644,"rel":645},"https:\u002F\u002Fnews.qq.com\u002Frain\u002Fa\u002F20251209A05V1G00",[638],[39,647,648,649],{},"AutoGLM arXiv 论文 ",[609,650,651],{"href":651,"rel":652},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2411.00820",[638],[39,654,655,656],{},"Chrome Web Store — 智谱清言 ChatGLM & AutoGLM 工作学习 AI 助手 ",[609,657,658],{"href":658,"rel":659},"https:\u002F\u002Fchromewebstore.google.com\u002Fdetail\u002Fmnpdbmgpebfihcndnpgdaihnkmloclkd",[638],[661,662,663],"style",{},"html pre.shiki code .sScJk, html code.shiki .sScJk{--shiki-default:#6F42C1;--shiki-dark:#B392F0}html pre.shiki code .sZZnC, html code.shiki .sZZnC{--shiki-default:#032F62;--shiki-dark:#9ECBFF}html pre.shiki code .sj4cs, html code.shiki .sj4cs{--shiki-default:#005CC5;--shiki-dark:#79B8FF}html pre.shiki code .sVt8B, html code.shiki .sVt8B{--shiki-default:#24292E;--shiki-dark:#E1E4E8}html pre.shiki code .sJ8bj, html code.shiki .sJ8bj{--shiki-default:#6A737D;--shiki-dark:#6A737D}html pre.shiki code .szBVR, html code.shiki .szBVR{--shiki-default:#D73A49;--shiki-dark:#F97583}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html.dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}",{"title":202,"searchDepth":250,"depth":250,"links":665},[666,667,668,669,670,671,672,673,674,675],{"id":23,"depth":224,"text":24},{"id":34,"depth":224,"text":34},{"id":108,"depth":224,"text":108},{"id":137,"depth":224,"text":138},{"id":194,"depth":224,"text":195},{"id":363,"depth":224,"text":363},{"id":520,"depth":224,"text":520},{"id":573,"depth":224,"text":574},{"id":603,"depth":224,"text":603},{"id":627,"depth":224,"text":627},"desktop","\u002Fimg\u002Ftools\u002Fautoglm.webp","AutoGLM 真实评测：智谱清言 + 清华大学合作的 GUI 自治智能体研究产品（arXiv 2411.00820）。2025-12-08 Open-AutoGLM 开源版本（MIT\u002FApache 2.0）发布，含 AutoGLM-Phone-9B 模型（HuggingFace \u002F ModelScope 可下载）+ Android 适配层 + 50+ 中文 App 任务示例。VAB-WebArena-Lite 55.2% 成功率（二次尝试 59.1%），中文 App 支持 Gmail\u002F微信\u002F淘宝\u002F美团等。",false,"md",[682,685,688,691],{"q":683,"a":684},"Open-AutoGLM 和原 AutoGLM 什么关系？","原 AutoGLM 是智谱 2024-10 发布的闭源整合产品（手机 + Web GUI agent），首次在真实手机环境跑通完整自治链路。Open-AutoGLM 是 2025-12 开源版本，把核心模型（AutoGLM-Phone-9B）+ 框架代码 + Android 适配层 + 50+ 中文 App 任务示例释放出来，方便开发者本地部署 + 二次开发。整合版闭源、商业；开源版用于学术 \u002F 私有化。",{"q":686,"a":687},"支持哪些 App？","Open-AutoGLM 内置 50+ 中文 App 示例：微信 \u002F 淘宝 \u002F 美团 \u002F 京东 \u002F 支付宝 \u002F 抖音 \u002F 小红书 \u002F 网易云 \u002F 大众点评 + 英文 Gmail \u002F Google Maps \u002F X \u002F Reddit 等。任何 Android App 理论可用，但效果取决于 UI 元素的可识别度，新版本 App 需要重新适配。",{"q":689,"a":690},"Web 端能力？","Web 端通过 Chrome 扩展（智谱清言：ChatGLM & AutoGLM 工作学习 AI 助手）提供——10 万+ 用户安装，含页面总结 \u002F 划线助手 \u002F 写作助手 \u002F 高级检索（知网 \u002F 知乎 \u002F 小红书）。VAB-WebArena-Lite benchmark 55.2% 首次成功率，二次尝试 59.1%。",{"q":692,"a":693},"和 Anthropic Computer Use \u002F Claude Desktop 怎么选？","AutoGLM 强在中文 App + 学术研究 + 开源权重 + 国内合规；Anthropic Computer Use 强在跨平台桌面 + 英文场景 + 模型推理；Claude Desktop 强在 MCP 生态 + 一键 .mcpb 安装。国内 App 自治 + 私有化 → AutoGLM；英文桌面 + 通用桌面操作 → Anthropic \u002F Claude Desktop。",[695,696],"zh","en",{},"\u002Ftools\u002Fagent\u002Fdesktop\u002Fautoglm","agent",[701,702,703],"android","chrome-extension","web",[705,709,713],{"plan":115,"price":706,"features":707,"notes":708},"$0","AutoGLM-Phone-9B 模型（MIT）+ 框架代码（Apache 2.0）+ Android 适配 + Remote ADB + 50+ 中文 App 示例","开源 \u002F 私有化部署",{"plan":710,"price":706,"features":711,"notes":712},"智谱清言 Chrome 扩展","Web 划线 \u002F 总结 \u002F 高级检索 \u002F 写作助手 \u002F 页面对话","Chrome Web Store",{"plan":714,"price":715,"features":716,"notes":717},"智谱清言 App 内置","Custom","完整 GUI agent 能力（Phone \u002F Web）+ 整合产品体验","国行 App \u002F 商业服务","Open-AutoGLM 开源免费（MIT 模型 + Apache 2.0 代码）\u002F 智谱清言 App 内置版商业","2026-06-19",[721],"onboarding\u002Fgui-agent-engineering",{"power":257,"ux":257,"price":264,"cn_support":264,"stability":250},{"title":12,"description":678},"agent\u002Fdesktop\u002Fautoglm",[726,729,731,733],{"name":727,"url":636,"accessed":728},"AutoGLM 项目主页（清华 \u002F Z.AI）","2026-06-24",{"name":730,"url":644,"accessed":728},"腾讯新闻 — 智谱开源 AutoGLM",{"name":732,"url":651,"accessed":728},"AutoGLM arXiv 论文 2411.00820",{"name":734,"url":658,"accessed":728},"Chrome Web Store — 智谱清言扩展（10万+用户）","tools\u002Fagent\u002Fdesktop\u002Fautoglm","智谱清言出品 + 清华团队 GUI 自治智能体——Open-AutoGLM 开源 9B 手机模型 + Chrome 浏览器扩展",[738,739,740,741,742,743,744],"gui-agent","mobile-agent","browser-agent","opensource","chinese","zhipu","chatglm","国内『AI 操作手机 \u002F 浏览器』开源标杆——智谱 32 个月研发、清华学术背景、MIT 开源模型权重。学术研究 \u002F 国内 App 自治 \u002F 私有化部署首选。生产稳定性 + 跨平台桌面仍弱于 Anthropic Computer Use \u002F Claude Desktop。","UpD6U6glOmQSbK7O4bjqwqdn7ZYE7JmmNZwkfE87YLc",{"id":748,"title":749,"alternatives":750,"api_compatible":755,"body":756,"category":1660,"chinese_friendly":257,"cover":1661,"description":1662,"domestic":679,"extension":680,"faq":6,"free":679,"github":6,"languages":1663,"meta":1664,"models":1665,"navigation":253,"notSuitable":1670,"opensource":679,"path":1675,"pillar":699,"platforms":1676,"priceTable":1680,"pricing":1696,"published":1697,"relatedPlaybooks":1698,"relatedReviews":1700,"score":1702,"self_host":679,"seo":1703,"seoTitle":6,"slug":1704,"sources":1705,"stem":1721,"suitable":1722,"tagline":1728,"tags":1729,"updated":728,"verdict":1736,"website":1145,"__hash__":1737},"tools\u002Ftools\u002Fagent\u002Fgeneral\u002Fmanus.md","Manus",[751,752,753,754],"agent\u002Fgeneral\u002Fgenspark","agent\u002Fgeneral\u002Felicit","agent\u002Fgeneral\u002Fdevin","agent\u002Fplatform\u002Fdify",[],{"type":18,"value":757,"toc":1640},[758,760,798,803,806,870,873,878,886,891,898,902,910,976,981,985,992,996,1004,1009,1026,1029,1032,1035,1042,1045,1051,1110,1116,1123,1128,1134,1138,1181,1184,1224,1227,1400,1405,1422,1427,1447,1450,1516,1518,1521,1538,1541,1562,1564,1614,1616,1633],[21,759,24],{"id":23},[761,762,767,783],"div",{"className":763},[764,765,766],"card","p-5","my-4",[26,768,769,772,773,776,777,782],{},[42,770,771],{},"一句话："," Butterfly Effect（中国创办、新加坡注册）2025-03-06 首发的通用 AI Agent，",[42,774,775],{},"邀请码一度被炒到 ¥5 万-10 万","（",[609,778,781],{"href":779,"rel":780},"https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FManus_%28AI_agent%29",[638],"据维基百科引用 China Daily 报道","）。多 sub-agent 并行架构——浏览、数据分析、代码执行、写作各自专门子 agent，主 orchestrator 自动路由到 Claude \u002F Qwen \u002F 自研模型，结果合成交付。",[26,784,785,786,776,789,793,794,797],{},"2025-12 传被 ",[42,787,788],{},"Meta 以约 20-30 亿美元收购",[609,790,792],{"href":779,"rel":791},[638],"Reuters \u002F AP 报道","），目前仍独立运营。最大价值是 ",[42,795,796],{},"deep research 类任务的引用密度和质量明显超过 ChatGPT Deep Research","。",[130,799,800],{},[26,801,802],{},"来源说明：本文基于 manus.im 官方页面、Wikipedia \"Manus (AI agent)\" 条目、第三方评测（Pick Right \u002F HyzenPro \u002F Info-Tech Research Group \u002F 36Kr \u002F 世界经济论坛企业页）综合整理。Meta 收购案细节、产品路线图仍在变化中，请以最新官方公告为准。",[21,804,805],{"id":805},"背景与公司",[36,807,808,819,825,831,837,848,858,864],{},[39,809,810,813,814,818],{},[42,811,812],{},"公司","：Butterfly Effect Pte. Ltd.（",[609,815,817],{"href":779,"rel":816},[638],"蝴蝶效应","），创始人 Xiao Hong（季逸超 Ji Yichao 为 Manus 联合创始人 + 首席科学家）",[39,820,821,824],{},[42,822,823],{},"公司位置","：办公室在北京 + 武汉 + 新加坡，目标市场北美 \u002F 日本 \u002F 韩国（不主打中国大陆）",[39,826,827,830],{},[42,828,829],{},"前作","：Monica，2023 年发布的浏览器扩展 AI 助手",[39,832,833,836],{},[42,834,835],{},"历史融资","：2024 年字节跳动曾出价 ~3000 万美元收购被拒（据 36Kr）",[39,838,839,842,843,847],{},[42,840,841],{},"Manus 启动","：2024-10 立项，灵感来自 ",[609,844,846],{"href":845},"\u002Fcoding\u002Fide\u002Fcursor.html","Cursor","；名字来自 MIT 拉丁校训 \"Mens et Manus\"（手与脑）",[39,849,850,853,854],{},[42,851,852],{},"首发数据","：2025-03-06 邀请制 beta，7 天内 200 万人候补，",[609,855,857],{"href":779,"rel":856},[638],"Demo 视频 20 小时百万播放",[39,859,860,863],{},[42,861,862],{},"营收","：2025-08 ARR 约 9000 万美元 → 2025-12 升至 1.25 亿美元",[39,865,866,869],{},[42,867,868],{},"Meta 收购","：2025-12-29 宣布，估值 20-30 亿美元，目前仍独立运营，但中国大陆已被屏蔽访问 + 关闭中文社交账号",[21,871,872],{"id":872},"核心特性",[874,875,877],"h3",{"id":876},"多-sub-agent-并行架构最大差异化","多 sub-agent 并行架构（最大差异化）",[26,879,880,885],{},[609,881,884],{"href":882,"rel":883},"https:\u002F\u002Fpick-right.com\u002Ftools\u002Fmanus-ai",[638],"Pick Right 2026-04 评测"," 描述的架构：",[130,887,888],{},[26,889,890],{},"\"Where most general-purpose AI products use one model end-to-end, Manus runs multiple specialized sub-agents in parallel: one handles web browsing, one handles data analysis, one handles code execution, one handles synthesis and writing.\"",[26,892,893,894,897],{},"主 orchestrator 给每一步选最合适的模型（Claude \u002F Qwen \u002F 自研），结果合成。这是 Manus 研究输出 ",[42,895,896],{},"引用密度更高、幻觉更少","的工程根源——分工 + 并行让每个 sub-agent 只做自己最擅长的部分。",[874,899,901],{"id":900},"通用-agent-模式","通用 Agent 模式",[26,903,904,909],{},[609,905,908],{"href":906,"rel":907},"https:\u002F\u002Fwww.infotech.com\u002Fresearch\u002Fassessing-manus-the-future-of-agentic-ai",[638],"Info-Tech 评测"," 列出的 GAIA 基准（General AI Assistants）：",[365,911,912,922],{},[368,913,914],{},[371,915,916,919],{},[374,917,918],{},"模型",[374,920,921],{},"GAIA 准确率",[389,923,924,936,944,952,960,968],{},[371,925,926,931],{},[394,927,928],{},[42,929,930],{},"Manus AI",[394,932,933],{},[42,934,935],{},">65%（SOTA）",[371,937,938,941],{},[394,939,940],{},"H2O.ai (h2oGPTe)",[394,942,943],{},"65%",[371,945,946,949],{},[394,947,948],{},"Google Langfun",[394,950,951],{},"49%",[371,953,954,957],{},[394,955,956],{},"Microsoft o1",[394,958,959],{},"38%",[371,961,962,965],{},[394,963,964],{},"OpenAI GPT-4o",[394,966,967],{},"32%",[371,969,970,973],{},[394,971,972],{},"OpenAI GPT-4 + Plugins",[394,974,975],{},"15-30%",[130,977,978],{},[26,979,980],{},"注：基准数据需第三方验证；Manus 官方公布数据，请审慎参考。",[874,982,984],{"id":983},"后台执行-长任务","后台执行 + 长任务",[26,986,987,988,991],{},"最特别的体验：任务下达后",[42,989,990],{},"可以关掉浏览器","，Manus 在云端继续跑（小时级），完成时通知。这种\"开着任务下班\"的模式是它 viral 的关键。",[874,993,995],{"id":994},"web-app-builder","Web App Builder",[26,997,998,999,1003],{},"直接生成完整网站和应用，内置数据库 + Stripe 支付 + SEO。但 ",[609,1000,1002],{"href":882,"rel":1001},[638],"Pick Right 评测"," 直白警告：",[130,1005,1006],{},[26,1007,1008],{},"\"Promising but buggy enough that I wouldn't ship to production from it yet.\"",[26,1010,1011,1012,1016,1017,1016,1021,1025],{},"复杂场景下出 bug 多，目前不建议生产部署，",[609,1013,1015],{"href":1014},"\u002Fcoding\u002Fbuilder\u002Fbolt-new.html","Bolt.new"," \u002F ",[609,1018,1020],{"href":1019},"\u002Fcoding\u002Fbuilder\u002Flovable.html","Lovable",[609,1022,1024],{"href":1023},"\u002Fcoding\u002Fbuilder\u002Fv0.html","v0"," 仍是 production app building 的更稳选项。",[874,1027,1028],{"id":1028},"桌面应用",[26,1030,1031],{},"提供 desktop app，能读本地文件 + 集成你的机器，不仅限于浏览器内任务。",[874,1033,1034],{"id":1034},"多模型路由",[26,1036,1037,1038,1041],{},"Claude \u002F Qwen \u002F Manus 自研模型，",[42,1039,1040],{},"按任务步骤自动选","——这是 Manus 跟单一模型 agent 的根本差异。",[21,1043,1044],{"id":1044},"价格与运行成本",[26,1046,1047,1050],{},[609,1048,1002],{"href":882,"rel":1049},[638]," 公开档位：",[365,1052,1053,1065],{},[368,1054,1055],{},[371,1056,1057,1060,1062],{},[374,1058,1059],{},"档位",[374,1061,108],{},[374,1063,1064],{},"关键点",[389,1066,1067,1077,1088,1099],{},[371,1068,1069,1072,1074],{},[394,1070,1071],{},"Free",[394,1073,706],{},[394,1075,1076],{},"每日有限 credits，够 1 个高强度任务\u002F天",[371,1078,1079,1082,1085],{},[394,1080,1081],{},"Pro",[394,1083,1084],{},"$20\u002F月",[394,1086,1087],{},"大多数付费用户落点；中等 credit + 多模型",[371,1089,1090,1093,1096],{},[394,1091,1092],{},"Plus",[394,1094,1095],{},"$50\u002F月",[394,1097,1098],{},"更高 credit + 优先队列，10+ 任务\u002F周",[371,1100,1101,1104,1107],{},[394,1102,1103],{},"Pro+ \u002F Team",[394,1105,1106],{},"最高 $200\u002F月",[394,1108,1109],{},"最大 credit + 团队空间（功能仍有限）",[26,1111,1112,1115],{},[42,1113,1114],{},"credit 经济学","：每个动作消耗 credits（浏览、代码运行、模型调用都计费）。",[26,1117,1118,1122],{},[609,1119,1121],{"href":882,"rel":1120},[638],"Pick Right 真实使用反馈","：",[130,1124,1125],{},[26,1126,1127],{},"\"Credits run out faster than the pricing page suggests. Heavy users routinely buy credit packs on top of subscriptions.\"",[26,1129,1130,1133],{},[42,1131,1132],{},"预算建议","：先用 Free 跑 3-5 个真实任务评估消耗速度，再决定档位。",[21,1135,1137],{"id":1136},"上手-5-分钟","上手 5 分钟",[629,1139,1140,1148,1151,1166,1169,1175,1178],{},[39,1141,1142,1143],{},"打开 ",[609,1144,1147],{"href":1145,"rel":1146},"https:\u002F\u002Fmanus.im",[638],"manus.im",[39,1149,1150],{},"账号注册（Google \u002F Apple OAuth 最快，国内手机号注册受限）",[39,1152,1153,1154,1157,1158],{},"给一个完整任务描述（",[42,1155,1156],{},"关键","：不要碎片化指令，给全场景）\n",[197,1159,1164],{"className":1160,"code":1162,"language":1163},[1161],"language-text","\"调研欧洲前 10 大 EV 充电网络（覆盖率、价格、可靠性、充电速度），\n产出带引用的 Markdown 对比表\"\n","text",[204,1165,1162],{"__ignoreMap":202},[39,1167,1168],{},"选模型路由（Auto 推荐）",[39,1170,1171,1172,1174],{},"提交后",[42,1173,990],{},"——任务在云端跑",[39,1176,1177],{},"完成后邮件 \u002F 站内通知",[39,1179,1180],{},"看结果 \u002F 下载交付物（Markdown \u002F Excel \u002F Word \u002F 网页）",[21,1182,1183],{"id":1183},"国内使用注意事项",[629,1185,1186,1196,1202,1208],{},[39,1187,1188,1122,1191,1195],{},[42,1189,1190],{},"大陆访问被屏蔽",[609,1192,1194],{"href":779,"rel":1193},[638],"Wikipedia 引用 36Kr 报道","，Butterfly Effect 已关闭中文社交账号、阻断中国大陆访问，原 Alibaba Qwen 合作版\"中文版 Manus\"已搁置",[39,1197,1198,1201],{},[42,1199,1200],{},"访问需稳定代理","：日韩 \u002F 美国节点",[39,1203,1204,1207],{},[42,1205,1206],{},"账号 \u002F 支付","：海外信用卡（Visa \u002F MasterCard），第三方代付方案有限",[39,1209,1210,1213,1214,1218,1219,1223],{},[42,1211,1212],{},"替代路径","：国内可考虑 ",[609,1215,1217],{"href":1216},"\u002Fagent\u002Fgeneral\u002Fgenspark.html","Genspark"," \u002F Devv \u002F ",[609,1220,1222],{"href":1221},"\u002Fagent\u002Fgeneral\u002Fflowith.html","元宝 Yuanbao"," \u002F 秘塔 Metaso 等",[21,1225,1226],{"id":1226},"与同类怎么选",[365,1228,1229,1253],{},[368,1230,1231],{},[371,1232,1233,1235,1237,1243,1247,1250],{},[374,1234,376],{},[374,1236,749],{},[374,1238,1239],{},[609,1240,1242],{"href":1241},"\u002Fcoding\u002Fagent\u002Fdevin.html","Devin",[374,1244,1245],{},[609,1246,1217],{"href":1216},[374,1248,1249],{},"ChatGPT Deep Research",[374,1251,1252],{},"GenAgent",[389,1254,1255,1275,1295,1314,1329,1345,1362,1380],{},[371,1256,1257,1260,1263,1266,1269,1272],{},[394,1258,1259],{},"核心定位",[394,1261,1262],{},"通用 Agent",[394,1264,1265],{},"AI 程序员",[394,1267,1268],{},"AI 搜索 + Agent",[394,1270,1271],{},"LLM Deep Research",[394,1273,1274],{},"通用",[371,1276,1277,1280,1283,1286,1289,1292],{},[394,1278,1279],{},"架构",[394,1281,1282],{},"多 sub-agent 并行",[394,1284,1285],{},"单 Agent + 沙盒",[394,1287,1288],{},"多模型",[394,1290,1291],{},"单模型",[394,1293,1294],{},"—",[371,1296,1297,1300,1303,1306,1309,1312],{},[394,1298,1299],{},"长任务",[394,1301,1302],{},"★★★★★ 小时级",[394,1304,1305],{},"★★★★★",[394,1307,1308],{},"★★★☆☆",[394,1310,1311],{},"★★★★☆",[394,1313,1308],{},[371,1315,1316,1319,1321,1323,1325,1327],{},[394,1317,1318],{},"引用密度",[394,1320,1305],{},[394,1322,1308],{},[394,1324,1311],{},[394,1326,1308],{},[394,1328,1308],{},[371,1330,1331,1334,1337,1339,1341,1343],{},[394,1332,1333],{},"App Builder",[394,1335,1336],{},"⚠️ 有但 buggy",[394,1338,425],{},[394,1340,425],{},[394,1342,425],{},[394,1344,425],{},[371,1346,1347,1350,1353,1356,1358,1360],{},[394,1348,1349],{},"中文",[394,1351,1352],{},"⚠️ 大陆屏蔽",[394,1354,1355],{},"⚠️",[394,1357,1305],{},[394,1359,1355],{},[394,1361,1311],{},[371,1363,1364,1366,1369,1372,1375,1378],{},[394,1365,108],{},[394,1367,1368],{},"$20-$200",[394,1370,1371],{},"$500\u002F月",[394,1373,1374],{},"$24.99\u002F月",[394,1376,1377],{},"随 ChatGPT Plus",[394,1379,1294],{},[371,1381,1382,1385,1388,1391,1394,1397],{},[394,1383,1384],{},"适合场景",[394,1386,1387],{},"research \u002F 数据分析",[394,1389,1390],{},"写代码 \u002F 修 bug",[394,1392,1393],{},"信息检索",[394,1395,1396],{},"单次深度研究",[394,1398,1399],{},"综合",[26,1401,1402,1122],{},[42,1403,1404],{},"选 Manus 如果你",[36,1406,1407,1413,1416,1419],{},[39,1408,1409,1410],{},"重视 research \u002F 数据分析任务的 ",[42,1411,1412],{},"引用密度和结构化输出",[39,1414,1415],{},"想试\"开任务下班、明早看结果\"的工作流",[39,1417,1418],{},"海外 \u002F 能解决账号网络问题",[39,1420,1421],{},"预算 $20-$50\u002F月，重度用户",[26,1423,1424,1122],{},[42,1425,1426],{},"别选 Manus 如果你",[36,1428,1429,1438,1441,1444],{},[39,1430,1431,1432,1016,1434,1437],{},"国内裸用（",[609,1433,1217],{"href":1216},[609,1435,1436],{"href":1221},"Yuanbao"," 更顺）",[39,1439,1440],{},"想生产部署 App（Web App Builder bug 多）",[39,1442,1443],{},"团队协作场景（功能不完善）",[39,1445,1446],{},"预算 \u003C $20\u002F月（Free 档够评估，付费档不一定划算）",[21,1448,1449],{"id":1449},"避坑清单",[36,1451,1452,1458,1468,1474,1480,1492,1498,1504],{},[39,1453,1454,1457],{},[42,1455,1456],{},"大陆访问已被官方屏蔽","：2025 年起阻断中国大陆访问，原\"中文版 Manus\"项目搁置",[39,1459,1460,1122,1463,1467],{},[42,1461,1462],{},"credit 烧得比官方页面暗示的快",[609,1464,1466],{"href":882,"rel":1465},[638],"Pick Right 2026 评测"," 真实反馈，重度用户经常额外买 credit pack",[39,1469,1470,1473],{},[42,1471,1472],{},"任务一启动无法控预算","：开始跑后只能 cancel 或等结束，credits 会一直消耗",[39,1475,1476,1479],{},[42,1477,1478],{},"Web App Builder 别上生产","：demo 漂亮，复杂场景翻车，Bolt.new \u002F Lovable \u002F v0 仍是生产部署更稳选项",[39,1481,1482,1485,1486,1491],{},[42,1483,1484],{},"每次任务从零开始","：没有持久 workspace，不像 ",[609,1487,1490],{"href":1488,"rel":1489},"https:\u002F\u002Fclaude.com",[638],"Claude Projects"," \u002F ChatGPT Custom GPTs 能跨会话记忆",[39,1493,1494,1497],{},[42,1495,1496],{},"没有 HubSpot \u002F Salesforce \u002F Notion \u002F Slack 集成","：拿到结果后要自己手动搬到工具栈",[39,1499,1500,1503],{},[42,1501,1502],{},"团队协作能力差","：单用户产品为主，无共享 workspace \u002F 评论 \u002F 审计",[39,1505,1506,1509,1510,1515],{},[42,1507,1508],{},"Meta 收购的不确定性","：Wikipedia 引用 ",[609,1511,1514],{"href":1512,"rel":1513},"https:\u002F\u002Fwww.nytimes.com\u002F2026\u002F03\u002F17\u002Ftechnology\u002Fchina-scrutiny-meta-manus.html",[638],"纽时 2026-03"," 报道，中国审查 Meta 收购案，长期路线图待观察",[21,1517,574],{"id":573},[26,1519,1520],{},"✅ 适合：",[36,1522,1523,1526,1529,1532,1535],{},[39,1524,1525],{},"委托多步骤 research（\"调研 X，产出 Markdown 对比表\"）",[39,1527,1528],{},"CSV \u002F Excel 重的数据分析任务",[39,1530,1531],{},"长任务 + 不想盯着看（小时级）",[39,1533,1534],{},"单兵作战的咨询顾问 \u002F 分析师 \u002F 创业者",[39,1536,1537],{},"对\"通用 Agent 体感天花板\"感兴趣的测试者",[26,1539,1540],{},"❌ 不适合：",[36,1542,1543,1546,1549,1552,1555],{},[39,1544,1545],{},"生产级应用部署",[39,1547,1548],{},"团队协作工作流",[39,1550,1551],{},"预算极敏感（free 档非常受限）",[39,1553,1554],{},"大陆稳定访问需求",[39,1556,1557,1558,1561],{},"需要深度持久上下文（用 ",[609,1559,1490],{"href":1488,"rel":1560},[638]," \u002F ChatGPT Custom GPT 等）",[21,1563,603],{"id":603},[36,1565,1566,1577,1588,1603],{},[39,1567,1568,1569,1016,1571,1573,1574,1576],{},"同类对比：",[609,1570,1242],{"href":1241},[609,1572,1217],{"href":1216}," \u002F Elicit \u002F ",[609,1575,1436],{"href":1221}," \u002F Metaso",[39,1578,1579,1580,1016,1584,1587],{},"概念：",[609,1581,1583],{"href":1582},"\u002Fwiki\u002Fai-agent.html","AI Agent",[609,1585,1586],{"href":1582},"Multi-Agent"," \u002F Computer Use \u002F Deep Research",[39,1589,1590,1591,1016,1595,1016,1599],{},"模型：",[609,1592,1594],{"href":1593},"\u002Fmodels\u002Fclaude-sonnet-4.html","Claude Sonnet 4",[609,1596,1598],{"href":1597},"\u002Fmodels\u002Fclaude-opus-4.html","Claude Opus 4",[609,1600,1602],{"href":1601},"\u002Fmodels\u002Fqwen-3.html","Qwen3",[39,1604,1605,1606,1016,1610],{},"进阶：",[609,1607,1609],{"href":1608},"\u002Fwiki\u002Fcontext-engineering.html","Context Engineering",[609,1611,1613],{"href":1612},"\u002Fwiki\u002Fprompt-engineering.html","Prompt Engineering",[21,1615,627],{"id":627},[36,1617,1618,1624,1627,1630],{},[39,1619,1620,1621],{},"官网：",[609,1622,1145],{"href":1145,"rel":1623},[638],[39,1625,1626],{},"Wikipedia：\"Manus (AI agent)\" 条目",[39,1628,1629],{},"第三方评测：pick-right.com \u002F hyzenpro.com \u002F infotech.com \u002F weforum.org",[39,1631,1632],{},"媒体报道：Reuters \u002F AP \u002F China Daily \u002F 36Kr \u002F NYT",[26,1634,1635,1636,1639],{},"本卡片由 AIHO 编辑部根据官方公开资料与第三方评测整理。所有事实点均标注来源；如发现价格 \u002F 功能 \u002F 公司状态与最新官方信息不一致，请通过 ",[609,1637,1638],{"href":1638},"\u002Fsubmit"," 反馈。",{"title":202,"searchDepth":250,"depth":250,"links":1641},[1642,1643,1644,1652,1653,1654,1655,1656,1657,1658,1659],{"id":23,"depth":224,"text":24},{"id":805,"depth":224,"text":805},{"id":872,"depth":224,"text":872,"children":1645},[1646,1647,1648,1649,1650,1651],{"id":876,"depth":250,"text":877},{"id":900,"depth":250,"text":901},{"id":983,"depth":250,"text":984},{"id":994,"depth":250,"text":995},{"id":1028,"depth":250,"text":1028},{"id":1034,"depth":250,"text":1034},{"id":1044,"depth":224,"text":1044},{"id":1136,"depth":224,"text":1137},{"id":1183,"depth":224,"text":1183},{"id":1226,"depth":224,"text":1226},{"id":1449,"depth":224,"text":1449},{"id":573,"depth":224,"text":574},{"id":603,"depth":224,"text":603},{"id":627,"depth":224,"text":627},"general","\u002Fimg\u002Ftools\u002Fmanus.webp","Manus 真实评测：Butterfly Effect（蝴蝶效应）出品的通用 AI Agent，多 sub-agent 并行架构 + 多模型路由（Claude \u002F Qwen \u002F 自研）。2025-03 首发即引爆，2025-12 传被 Meta 以约 20 亿美元收购。AIHO 编辑部基于官方文档与多份评测整理。",[695,696],{},[1666,1667,1668,1669],"claude-sonnet-4","claude-opus-4","qwen-max","manus-internal",[1671,1672,1673,1674],"生产级 Web App 部署（Web App Builder 仍有 bug）","团队协作场景（功能未完善）","需要持久工作空间 \u002F 跨会话上下文","对成本极敏感（credits 烧得快）","\u002Ftools\u002Fagent\u002Fgeneral\u002Fmanus",[703,1677,1678,1679],"windows","macos","linux",[1681,1684,1688,1691],{"plan":1071,"price":706,"limit":1682,"cn_pay":1294,"note":1683},"每日有限 credits，足够每天 1 个 demanding 任务","试水\u002F评估",{"plan":1081,"price":1084,"limit":1685,"cn_pay":1686,"note":1687},"更高 credit + 更长任务时长 + 多模型路由","⚠️ 需海外卡","个人主力档",{"plan":1092,"price":1095,"limit":1689,"cn_pay":1355,"note":1690},"更大 credit 池 + 优先队列","10+ 任务\u002F周",{"plan":1692,"price":1693,"limit":1694,"cn_pay":1355,"note":1695},"Pro+\u002FTeam","$200\u002F月","最大 credit + 团队工作空间（功能受限）","团队功能仍有限","Free \u002F Pro $20\u002Fmo \u002F Plus $50\u002Fmo \u002F Pro+ Team 最高 $200\u002Fmo","2026-06-18",[1699],"onboarding\u002Fmanus-getting-started",[1701],"manus-deep-review",{"power":264,"ux":264,"price":250,"cn_support":257,"stability":250},{"title":749,"description":1662},"agent\u002Fgeneral\u002Fmanus",[1706,1708,1711,1713,1715,1718],{"title":1707,"url":1145},"Manus 官网",{"title":1709,"url":1710},"Manus 维基百科","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FManus_(AI_agent)",{"title":1712,"url":882},"Manus AI Review 2026 (Pick Right)",{"title":1714,"url":906},"Info-Tech Manus Assessment",{"title":1716,"url":1717},"WEF Butterfly Effect Page","https:\u002F\u002Fwww.weforum.org\u002Forganizations\u002Fbutterfly-effect",{"title":1719,"url":1720},"HyzenPro 2026 Review","https:\u002F\u002Fhyzenpro.com\u002Fblog\u002Fmanus-ai-review","tools\u002Fagent\u002Fgeneral\u002Fmanus",[1723,1724,1725,1726,1727],"需要委托多步骤研究 \u002F 报告（带引用密度的 deep research）","数据分析（CSV \u002F Excel 重的任务）","需要","调研类工作（市场调研、竞品对比、文献综述）","对 Computer Use 类 Agent 实践有兴趣的人","通用 Agent 体感天花板，自主完成复杂多步骤任务",[1730,1731,1732,1733,1734,1735],"general-agent","autonomous","computer-use","multi-agent","butterfly-effect","monica","通用 Agent 体感天花板代表。多 sub-agent 并行 + 自动模型路由让 research \u002F 数据分析输出明显比 ChatGPT Deep Research 引用密度更高。慢、credit 烧得快、Web App Builder 仍有 bug，但任务跑通时确实让人惊艳。","AaSPUCWb2nxYWHxSC_BBzBy3OJG69bjyadYk9bISuwc",1783173059599]