[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-77922":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":9,"language":10,"languages":9,"totalLinesOfCode":9,"stars":11,"forks":12,"watchers":13,"openIssues":14,"contributorsCount":15,"subscribersCount":15,"size":15,"stars1d":15,"stars7d":16,"stars30d":17,"stars90d":15,"forks30d":15,"starsTrendScore":18,"compositeScore":19,"rankGlobal":9,"rankLanguage":9,"license":20,"archived":21,"fork":21,"defaultBranch":22,"hasWiki":23,"hasPages":21,"topics":24,"createdAt":9,"pushedAt":9,"updatedAt":25,"readmeContent":26,"aiSummary":27,"trendingCount":15,"starSnapshotCount":15,"syncStatus":18,"lastSyncTime":28,"discoverSource":29},77922,"openteam","afumu\u002Fopenteam","afumu","A Chrome extension for orchestrating local AI-agent team workflows across web AI services.",null,"TypeScript",151,41,1,19,0,6,104,2,4.87,"Other",false,"main",true,[],"2026-06-12 02:03:45","# OpenTeam\n\n**🌐 Language:** English | [简体中文](README.zh-CN.md)\n\n> A local-first AI team workspace that turns your existing ChatGPT, Claude, Gemini, DeepSeek, and Grok web sessions into a multi-agent discussion room.\n\nOpenTeam is a Manifest V3 Chrome extension. It does not require model API keys and does not spend extra OpenAI, Claude, Gemini, DeepSeek, or Grok API tokens. Instead, it reuses the AI accounts you already have open in your browser, sends tasks to those web sessions, and gathers replies from different people and models into one shared team chat.\n\nUse it for learning, research, and personal non-commercial experiments where one answer is not enough: product reviews, technical design reviews, content planning, personal decisions, multi-model comparison, and multi-step work that benefits from several AI roles thinking together.\n\n## 🌱 Background\n\nOpenTeam is a sister project to [OpenLink](https:\u002F\u002Fgithub.com\u002Fafumu\u002Fopenlink), an earlier browser extension I built around using the browser itself as the working surface. While building OpenLink, I already wanted a companion extension that could reuse signed-in browser sessions to talk with large language models and compare answers across models, but that idea stayed on the shelf while other work got busy.\n\nThe idea became concrete after I met Lu ([YUANLU007](https:\u002F\u002Fgithub.com\u002FYUANLU007)), a news editor. In her writing workflow, she often needs to compare several AI systems, inspect how different models answer the same question, and use multiple perspectives for factual checks. That matched the original OpenLink idea closely, so OpenTeam became its sister project: a local browser workspace for sending one task to multiple AI web sessions and gathering their responses in one discussion room.\n\n![OpenTeam group chat preview](docs\u002Fassets\u002Fgroup-chat-ui-concept.png)\n\n## 💬 Community\n\nJoin the OpenTeam WeChat community by scanning the QR code below. If the QR code has expired, add WeChat ID `afumudev` and include `OpenTeam` in your request so I can invite you to the group.\n\n\u003Cp>\n  \u003Cimg src=\"docs\u002Fassets\u002Fopenteam-wechat-group-qr.jpg\" alt=\"OpenTeam WeChat community QR code\" width=\"280\">\n\u003C\u002Fp>\n\n## ✨ Highlights\n\n- 🚫 **0 API token workflow**: reuse AI website sessions instead of calling model APIs directly.\n- 🧩 **Multi-model discussion**: coordinate Gemini, ChatGPT, Claude, DeepSeek, Grok, and other supported web sessions in one chat.\n- 🧑‍🏫 **Built-in advisor library**: start with 38 expert and thinking-style advisor templates, or create your own people.\n- 📣 **Mention-based routing**: use `@person` to ask one member, or `@everyone` to dispatch the same task to the whole team.\n- 🔄 **Independent and collaborative modes**: compare isolated perspectives first, then let members reference, challenge, and build on each other.\n- 💾 **Local-first storage**: chats, people, messages, notes, highlights, and settings are stored in browser storage.\n- 🤖 **Agent control CLI**: optional `openteamcli` support lets local agents create chats, add roles, post tasks, and wait for replies.\n\n## 🧭 How It Works\n\nEach OpenTeam member is bound to an AI website session. When you send a message, OpenTeam builds a prompt from the chat mode, member persona, referenced messages, and shared context. The extension then delivers that prompt into the member's iframe-backed AI page and listens for the reply.\n\n```text\nteam.html\n  -> background service worker\n  -> AI site iframe\n  -> content script\n  -> AI webpage reply\n  -> OpenTeam message stream\n```\n\nSupported site types:\n\n| Site | Common fit |\n| --- | --- |\n| Gemini | Long context, research, multimodal material |\n| ChatGPT | General execution, tool-shaped workflows, fast iteration |\n| Claude | Long documents, review, structured writing, careful reasoning |\n| DeepSeek | Code, reasoning, Chinese-language work, cost-sensitive tasks |\n| Grok | Current-events research, alternative phrasing, X\u002FGrok web-session workflows |\n\nThese are practical defaults, not rankings. Actual quality depends on model versions, account plans, task type, and input material.\n\n## ⚠️ Disclaimer\n\nOpenTeam is an unofficial project for learning, research, and personal non-commercial use only. It is not affiliated with, endorsed by, or supported by OpenAI, Anthropic, Google, DeepSeek, xAI, or any supported AI website.\n\nOpenTeam interacts with AI websites through user-authenticated browser sessions and DOM automation. Website changes, account rules, rate limits, anti-abuse systems, or terms of service may affect whether it works. You are responsible for reviewing and complying with the rules, policies, laws, and regulations that apply to your use.\n\nDo not use OpenTeam for commercial products, hosted services, paid workflows, bulk automation, spam, scraping, bypassing access controls, bypassing paid access, or any activity that violates third-party terms or rights.\n\nUse it at your own risk. The maintainers are not responsible for account restrictions, service interruptions, data loss, legal issues, or other consequences arising from your use. This notice is not legal advice.\n\n## 🚀 Install\n\n### Option 1: Build the Extension From Source\n\nMost users only need the browser extension:\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Fafumu\u002Fopenteam.git\ncd openteam\nnpm install\nnpm run build\n```\n\nAfter the build finishes, open `chrome:\u002F\u002Fextensions\u002F`, turn on Developer mode, click **Load unpacked**, select the generated `dist\u002F` directory, and then click the OpenTeam extension icon to open the team workspace.\n\n### Option 2: CLI + Agent Skill\n\nUse this path if you want Codex, Claude Code, or another local agent to control OpenTeam chats. Install the CLI from npm, then install the agent skill:\n\n```bash\nnpm install -g @afumu\u002Fopenteamcli\nnpx skills add afumu\u002Fopenteam --skill openteam-control\nopenteamcli daemon start\nopenteamcli doctor\n```\n\nThe skills installer will ask which agent, scope, and install method to use. Use the GitHub command above for the public repository; if you are developing from a local checkout or unzipped source package, install the skill from the repository root instead:\n\n```bash\nnpx skills add . --skill openteam-control\n```\n\nAfter installation, open the OpenTeam extension page and enable local agent control in settings. See [OpenTeam CLI](packages\u002Fopenteamcli\u002FREADME.md) for more CLI details.\n\n## 🛠️ Development From Source\n\n```bash\nnpm install\nnpm run dev\n```\n\nBuild the extension and load the generated `dist\u002F` directory from `chrome:\u002F\u002Fextensions\u002F` with Developer mode enabled.\n\nUseful checks:\n\n```bash\nnpm run typecheck\nnpm test\nnpm run build\nnpm run verify\n```\n\n`npm run verify` runs type checking, unit tests, and a production build.\n\n## 🔐 Permissions and Privacy\n\nOpenTeam is local-first, but it needs browser-extension permissions that are worth reviewing before installation:\n\n- `storage`: save chats, people, notes, settings, and local state.\n- `tabs`: locate and communicate with extension and AI-site tabs\u002Fframes.\n- `alarms`: schedule runtime maintenance tasks.\n- `declarativeNetRequest`: adjust response headers so supported AI websites can be embedded in the extension iframe workspace.\n- `clipboardRead` \u002F `clipboardWrite`: support copy and clipboard-based interactions.\n- Host permissions: allow OpenTeam to work with supported AI websites and embedded web sessions.\n\nOpenTeam does not provide cloud sync. Your AI conversations are still processed by the AI websites you use, under their own account rules, usage limits, privacy policies, and terms of service.\n\n## 🚧 Current Limits\n\n- OpenTeam is Chrome \u002F Chromium-first.\n- AI-site adapters depend on page DOM structure, so website redesigns can break prompt sending or reply capture.\n- The iframe workspace requires response-header changes through `declarativeNetRequest`, which makes the permission surface heavier than a normal popup extension.\n- Built-in famous-person advisor templates are prompt templates inspired by public ideas. They are not real people and should not be presented as real participation.\n- Medical, legal, financial, and other high-stakes outputs still require human judgment and qualified professional advice.\n\n## 🗂️ Repository Layout\n\n```text\npublic\u002F                 Chrome extension manifest, team page, styles, DNR rules\nsrc\u002Fbackground\u002F         service worker, command handlers, runtime routing\nsrc\u002Fcontent\u002F            AI-site content scripts, adapters, reply observation\nsrc\u002Fgroup\u002F              group data model, store, roles, prompts, mention parsing\nsrc\u002FteamPage\u002F           OpenTeam workspace UI\npackages\u002Fopenteamcli\u002F   local CLI and daemon for agent control\ndocs\u002F                   design documents and assets\n```\n\n## 🤝 Contributing\n\nIssues and pull requests are welcome. Good starting areas include:\n\n- AI-site adapter fixes when target websites change.\n- Permission and privacy hardening.\n- Test coverage for group routing, prompt construction, storage, and UI flows.\n- Documentation, examples, and onboarding improvements.\n- New advisor templates or orchestration patterns with clear boundaries.\n\nBefore opening a pull request, run:\n\n```bash\nnpm run verify\n```\n\n## 📚 Documentation\n\n- [Design document](docs\u002FDESIGN.md)\n- [OpenTeam CLI](packages\u002Fopenteamcli\u002FREADME.md)\n\n## 📜 License\n\nOpenTeam is released under the [PolyForm Noncommercial License 1.0.0](LICENSE), using the SPDX identifier `PolyForm-Noncommercial-1.0.0`.\n\nYou may use, study, modify, and redistribute the project for non-commercial purposes only. Commercial use, commercial redistribution, hosted commercial services, paid workflows, and productized commercial use are not permitted without separate written permission.\n","OpenTeam 是一个 Chrome 扩展程序，旨在通过整合本地 AI 代理团队工作流程来协调网页上的多种 AI 服务。其核心功能包括零 API 令牌工作流、多模型讨论室以及基于提及的路由机制等，使得用户可以利用已登录的 ChatGPT、Claude、Gemini 等 AI 会话进行任务分配，并将不同来源的回答汇集到同一个聊天界面中。此外，该工具还提供了内置顾问库与独立协作模式切换等功能，增强了团队讨论的灵活性和深度。适用于需要从多个角度获取信息或见解的学习研究场景、产品评价、技术设计评审、内容规划及个人决策等领域。","2026-06-11 03:56:12","CREATED_QUERY"]