WebBrain 作者: Emre Sokullu
Open-source AI browser agent — chat with pages, automate tasks, multi-provider LLM support.
113 位使用者113 位使用者
擴充套件後設資料
畫面擷圖
關於此擴充套件
WebBrain is a free, open-source browser extension that brings AI agent capabilities to Chrome. Read pages, extract data, and automate web tasks, powered by your choice of LLM. The self-hostable alternative to proprietary browser AI plugins.
And when the network goes away, it keeps working.
Key Features:
- Page Understanding: Reads and comprehends any web page, from articles and docs to dashboards and forms
- Full Browser Agent: Click, type, scroll, navigate, and interact with pages on your behalf
- Data Extraction: Pull structured data off any page, including tables, lists, links, and forms
- Multi-Provider LLM: Works with local LLMs (LM Studio, Ollama, llama.cpp) and leading cloud providers (OpenAI, Anthropic Claude, Gemini, OpenRouter, and more)
- Privacy First: Your data stays yours. Run with a local LLM for zero data leakage
- Smart Context: Automatic context management prevents token overflow
Apocalypse Mode:
Optional and off by default. Turn it on and WebBrain installs what it needs to answer questions with no internet at all: an on-device text model, an on-device vision model, and a Wikipedia archive in the openZIM format used by Kiwix.
Local models need a browser with WebGPU. Archives range from a few hundred MB to tens of GB, so WebBrain shows the exact size, source, and license and asks before downloading anything.
Interaction Modes:
- Ask Mode: Read-only questions about current page
- Act Mode: Full agent with browser control for automation
License: MIT License
Developer: Emre Sokullu
GitHub: https://github.com/webbrain-one/webbrain
Categories: Productivity, Developer Tools, AI/Machine Learning
For the full changelog, check out https://github.com/webbrain-one/webbrain/blob/main/CHANGELOG.md
And when the network goes away, it keeps working.
Key Features:
- Page Understanding: Reads and comprehends any web page, from articles and docs to dashboards and forms
- Full Browser Agent: Click, type, scroll, navigate, and interact with pages on your behalf
- Data Extraction: Pull structured data off any page, including tables, lists, links, and forms
- Multi-Provider LLM: Works with local LLMs (LM Studio, Ollama, llama.cpp) and leading cloud providers (OpenAI, Anthropic Claude, Gemini, OpenRouter, and more)
- Privacy First: Your data stays yours. Run with a local LLM for zero data leakage
- Smart Context: Automatic context management prevents token overflow
Apocalypse Mode:
Optional and off by default. Turn it on and WebBrain installs what it needs to answer questions with no internet at all: an on-device text model, an on-device vision model, and a Wikipedia archive in the openZIM format used by Kiwix.
- Offline Wikipedia: Pick a language and edition, from a compact Simple English archive to a full edition with images, and read articles in a built-in reader
- Emergency Box: A field library of curated PDFs covering OpenStax textbooks, emergency health guides, and field manuals, each downloaded only when you ask for it
- Offline Answer Engine: Local search across your installed archives, so answers quote the passages they came from and link back to a source you can open and check
- Emergency Communication: A word board covering 110 everyday concepts across 1,756 languages, built from the public-domain PanLex Swadesh Corpus, for when you need to be understood without a shared language
- Stays on your device: Once a component is installed, questions, retrieval, and answers all run locally
Local models need a browser with WebGPU. Archives range from a few hundred MB to tens of GB, so WebBrain shows the exact size, source, and license and asks before downloading anything.
Interaction Modes:
- Ask Mode: Read-only questions about current page
- Act Mode: Full agent with browser control for automation
License: MIT License
Developer: Emre Sokullu
GitHub: https://github.com/webbrain-one/webbrain
Categories: Productivity, Developer Tools, AI/Machine Learning
For the full changelog, check out https://github.com/webbrain-one/webbrain/blob/main/CHANGELOG.md
由 1 位評論者給出 5 分
權限與資料
必要權限:
- 取得剪貼簿中的資料
- 輸入資料到剪貼簿
- 下載檔案、讀取或修改瀏覽器的下載紀錄
- 向您顯示通知
- 存取瀏覽器分頁
- 在上網時了解瀏覽器行為狀態
- 存取您所有網站中的資料
收集下列資料:
- 開發者聲稱此擴充套件不要求收集任何資料。
開發者聲稱選擇性收集下列資料:
- 驗證資訊
- 私人通訊內容
- 可識別個人資訊
- 技術與互動資料
- 網站內容
更多資訊
- 版本
- 34.1.0
- 大小
- 13.63 MB
- 最近更新
- 2 天前 (2026年9月1日)
- 授權條款
- MIT License
- 隱私權保護政策
- 閱讀此附加元件的隱私權保護政策
- 版本紀錄
- 新增至收藏集