chore(version): bump version to 1.7.1

* docs(page): lm studio doc
* docs: update documentation and improve model section

.github/workflows/deploy-demo.yml

@@ -28,7 +28,7 @@ jobs:
         run: npm run build:website
 
       - name: Setup Pages
-        uses: actions/configure-pages@v5
+        uses: actions/configure-pages@v6
 
       - name: Upload artifact
         uses: actions/upload-pages-artifact@v4
@@ -37,4 +37,4 @@ jobs:
 
       - name: Deploy to GitHub Pages
         id: deployment
-        uses: actions/deploy-pages@v4
+        uses: actions/deploy-pages@v5

CONTRIBUTING.md

@@ -2,42 +2,13 @@
 
 ♥️ We welcome contributions from everyone.
 
-## 🚀 Quick Start
-
-### Development Setup
-
-1. **Prerequisites**
-    - `macOS` / `Linux` / `WSL`
-    - `node.js >= 20` with `npm >= 10`
-    - An editor that supports `ts/eslint/prettier`
-    - Make sure `eslint`, `prettier` and `commitlint` work well. Un-linted code won't pass the CI.
-
-2. **Setup**
-
-    ```bash
-    npm i
-    npm start          # Start demo and documentation site
-    npm run build      # Build libs and website
-    ```
-
-### Project Structure
-
-This is a **monorepo** with npm workspaces containing **4 main packages**:
-
-- **Page Agent** (`packages/page-agent/`) - Main entry with built-in UI Panel, published as `page-agent` on npm
-- **Core** (`packages/core/`) - Core agent logic without UI (npm: `@page-agent/core`)
-- **Extension** (`packages/extension/`) - Chrome extension for multi-page tasks and browser-level automation
-- **Website** (`packages/website/`) - React documentation and landing page. Also as demo and test page for the core lib. private package `@page-agent/website`
-
-> We use a simplified monorepo solution with `native npm-workspace + ts reference + vite alias`. No fancy tooling. Hoisting is required.
-> 
-> - When developing. Use alias so that we don't have to pre-build.
-> - When bundling. Use external and disable ts `paths` alias.
-> - When bundling `IIFE` and `Website`. Bundle everything together.
+For local development workflows, setup, local LLM config, extension development, testing on other websites, and more details, see [docs/developer-guide.md](docs/developer-guide.md).
 
 ## 🤝 How to Contribute
 
-### Reporting Issues
+> **[Maintainer's Note](https://github.com/alibaba/page-agent/issues/349)**
+
+### Opening Issues
 
 - Use the GitHub issue tracker to report bugs or request features
 - Search existing issues before creating new ones
@@ -46,147 +17,24 @@ This is a **monorepo** with npm workspaces containing **4 main packages**:
 
 ### Code Contributions
 
-1. **Fork and Clone**
-
-    ```bash
-    git clone https://github.com/your-username/page-agent.git
-    cd page-agent
-    ```
-
-2. **Create Feature Branch**
-
-    ```bash
-    git checkout -b feat/your-feature-name
-    ```
-
-3. **Make Changes**
-    - Follow existing code style and patterns
-    - Add tests for new functionality
-    - Update documentation as needed
-
-4. **Test Your Changes**
-    - Build and lint everything.
-    - Test in our demo website
-    - Test it on other websites if applicable
-    - `@TODO: test suite`
-
-5. **Commit and Push**
-
-    ```bash
-    git add .
-    git commit -m "feat: add awesome feature"
-    git push origin feat/your-feature-name
-    ```
-
-6. **Create Pull Request**
-    - Provide clear description of changes
-    - Link related issues
-    - Include screenshots for UI changes
-
-## 📝 Code Style
-
-### General Guidelines
-
-- Use TypeScript for type safety
-- Follow existing naming conventions
-- Write meaningful commit messages
-- Keep functions small and focused
-- Add JSDoc for public APIs
+1. Follow existing code style and patterns
+2. Update documentation as needed
+3. Add JSDoc for public APIs
+4. Build and lint everything
+5. Test in our demo website, and on other websites if applicable
+6. Include screenshots for UI changes
 
 ### Vibe Coding with AI
 
-> [Vibe coding](https://en.wikipedia.org/wiki/Vibe_coding)
-
+- Vibe coding is **NOT** allowed for the core lib or the extension!!!
 - Vibe coding is **RECOMMENDED** when maintaining **the demo, the website, the UI and tests**.
-    - We have a [website/AGENTS.md](packages/website/AGENTS.md) for that.
-- Vibe coding is **NOT** allowed for the core lib!!!
-- NEVER try to vibe coding the MV3 extension!!! It is HELL.
+- Make sure your AI references `AGENTS.md` and `website/AGENTS.md` for better quality.
 - Review anything AI wrote before make a commit. You are the author of anything you commit. NOT AI.
 
-If your AI assistant does not support [AGENTS.md](https://agents.md/). Add an alias for it:
-
-- claude-code (`CLAUDE.md`)
-
-    ```markdown
-    @AGENTS.md
-    ```
-
-- antigravity (`.agent/rules/alias.md`)
-
-    ```markdown
-    ---
-    trigger: always_on
-    ---
-
-    @../../AGENTS.md
-    ```
-
-## 🔧 Development Workflows
-
-### Test With Your Own LLM API
-
-- Create a `.env` file in the repo root with your LLM API config
-
-    ```env
-    LLM_MODEL_NAME=gpt-5.2
-    LLM_API_KEY=your-api-key
-    LLM_BASE_URL=https://api.your-llm-provider.com/v1
-    ```
-
-- **Ollama example** (tested on 0.15 + qwen3:14b, RTX3090 24GB):
-
-    ```env
-    LLM_BASE_URL="http://localhost:11434/v1"
-    LLM_API_KEY="NA"
-    LLM_MODEL_NAME="qwen3:14b"
-    ```
-
-    > @see https://alibaba.github.io/page-agent/docs/features/models#ollama for configuration
-
-- **Restart the dev server** to load new env vars
-- If not provided, the demo will use the free testing proxy by default. By using it, you agree to its [terms](https://github.com/alibaba/page-agent/blob/main/docs/terms-and-privacy.md).
-
-### Extension Development
-
-```bash
-# make sure you ran `npm run build:libs` first
-# and every time you changed the core libs
-npm run dev -w @page-agent/ext
-npm run zip -w @page-agent/ext
-```
-
-- Update `packages/extension/docs/extension_api.md` for API integration details
-
-### Testing on Other Websites
-
-- Start and serve a local `iife` script
-
-    ```bash
-    npm run dev:demo # Serving IIFE with auto rebuild at http://localhost:5174/page-agent.demo.js
-    ```
-
-- Add a new bookmark
-
-    ```javascript
-    javascript:(function(){var s=document.createElement('script');s.src=`http://localhost:5174/page-agent.demo.js?t=${Math.random()}`;s.onload=()=>console.log(%27PageAgent ready!%27);document.head.appendChild(s);})();
-    ```
-
-- Click the bookmark on any page to load Page-Agent
-
-> Warning: AK in your local `.env` will be inlined in the iife script. Be very careful when you distribute the script.
-
-### Adding Documentation
-
-Ask an AI to help you add documentation to the `website/` package. Follow the existing style.
-
-> Our AGENTS.md file and guardrails are designed for this purpose. But please be careful and review anything AI generated.
-
 ## 🚫 What We Don't Accept
 
 - Breaking changes and large PRs without prior discussion
 - Heavy dependencies to core libs
-- Contributions without proper testing
-- Code that doesn't follow project conventions
 - Dependencies or code with licenses incompatible with MIT
 - Bot or AI-generated pull requests without meaningful human involvement
 
@@ -194,12 +42,6 @@ Ask an AI to help you add documentation to the `website/` package. Follow the ex
 
 By contributing to this project, you agree that your contributions will be licensed under the MIT License.
 
-> CLA is optional.
-
-## 💬 Questions?
-
-- Open a GitHub issue for technical questions
-- Check existing documentation and issues first
-- Be respectful and constructive in discussions
+---
 
 Thank you for helping make PageAgent better! 🎉

README.md

@@ -25,15 +25,16 @@ The GUI Agent Living in Your Webpage. Control web interfaces with natural langua
 - **📖 Text-based DOM manipulation**
     - No screenshots. No multi-modal LLMs or special permissions needed.
 - **🧠 Bring your own LLMs**
-- **🎨 Pretty UI with human-in-the-loop**
 - **🐙 Optional [chrome extension](https://alibaba.github.io/page-agent/docs/features/chrome-extension) for multi-page tasks.**
+    - And an [MCP Server (Beta)](https://alibaba.github.io/page-agent/docs/features/mcp-server) to control it from outside
 
 ## 💡 Use Cases
 
 - **SaaS AI Copilot** — Ship an AI copilot in your product in lines of code. No backend rewrite.
 - **Smart Form Filling** — Turn 20-click workflows into one sentence. Perfect for ERP, CRM, and admin systems.
 - **Accessibility** — Make any web app accessible through natural language. Voice commands, screen readers, zero barrier.
-- **Multi-page Agent** — Extend your own agent's reach across browser tabs with the optional [chrome extension](https://alibaba.github.io/page-agent/docs/features/chrome-extension).
+- **Multi-page Agent** — Extend your own web agent's reach across browser tabs [chrome extension](https://alibaba.github.io/page-agent/docs/features/chrome-extension).
+- **MCP** - Allow your agent clients to control your browser.
 
 ## 🚀 Quick Start
 
@@ -49,8 +50,8 @@ Fastest way to try PageAgent with our free Demo LLM:
 
 | Mirrors | URL                                                                                |
 | ------- | ---------------------------------------------------------------------------------- |
-| Global  | https://cdn.jsdelivr.net/npm/page-agent@1.5.9/dist/iife/page-agent.demo.js         |
-| China   | https://registry.npmmirror.com/page-agent/1.5.9/files/dist/iife/page-agent.demo.js |
+| Global  | https://cdn.jsdelivr.net/npm/page-agent@1.7.1/dist/iife/page-agent.demo.js         |
+| China   | https://registry.npmmirror.com/page-agent/1.7.1/files/dist/iife/page-agent.demo.js |
 
 ### NPM Installation
 
@@ -73,13 +74,27 @@ await agent.execute('Click the login button')
 
 For more programmatic usage, see [📖 Documentations](https://alibaba.github.io/page-agent/docs/introduction/overview).
 
+## 🌟 Awesome Page Agent
+
+Built something cool with PageAgent? Add it here! Open a PR to share your project.
+
+> These are community projects — not maintained or endorsed by us. Use at your own discretion.
+
+| Project  | Description                                                 |
+| -------- | ----------------------------------------------------------- |
+| _Yours?_ | [Open a PR](https://github.com/alibaba/page-agent/pulls) 🙌 |
+
 ## 🤝 Contributing
 
-We welcome contributions from the community! Follow our instructions in [CONTRIBUTING.md](CONTRIBUTING.md) for setup and guidelines.
+We welcome contributions from the community! See [CONTRIBUTING.md](CONTRIBUTING.md) for guidelines and [docs/developer-guide.md](docs/developer-guide.md) for local development workflows.
 
-Please read [Code of Conduct](docs/CODE_OF_CONDUCT.md) before contributing.
+Please read the [maintainer's note](https://github.com/alibaba/page-agent/issues/349) on principles and current state.
 
-Contributions generated entirely by bots or agents without substantial human involvement will not be accepted, and bot accounts may be blocked.
+Contributions generated entirely by **bots or AI** without substantial human involvement will **not be accepted**.
+
+## ⚖️ License
+
+[MIT License](LICENSE)
 
 ## 👏 Acknowledgments
 
@@ -97,23 +112,8 @@ Licensed under the MIT License
 We gratefully acknowledge the browser-use project and its contributors for their
 excellent work on web automation and DOM interaction patterns that helped make
 this project possible.
-
-Third-party dependencies and their licenses can be found in the package.json
-file and in the node_modules directory after installation.
 ```
 
-## 📄 License
-
-[MIT License](LICENSE)
-
 ---
 
 **⭐ Star this repo if you find PageAgent helpful!**
-
-<a href="https://www.star-history.com/?repos=alibaba%2Fpage-agent&type=date">
- <picture>
-   <source media="(prefers-color-scheme: dark)" srcset="https://api.star-history.com/image?repos=alibaba/page-agent&type=date&theme=dark&legend=top-left&v=7" />
-   <source media="(prefers-color-scheme: light)" srcset="https://api.star-history.com/image?repos=alibaba/page-agent&type=date&legend=top-left&v=7" />
-   <img alt="Star History Chart" src="https://api.star-history.com/image?repos=alibaba/page-agent&type=date&legend=top-left&v=7" />
- </picture>
-</a>

docs/CHANGELOG.md

@@ -5,6 +5,31 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [1.6.0] - 2026-03-21
+
+### Features
+
+- **Beta MCP support** - New `@page-agent/mcp` package lets MCP clients such as Claude Desktop and Copilot control the browser through the Page Agent extension
+- **Better iframe handling** - Same-origin iframe elements are handled more reliably during DOM extraction and actions
+- **Extension history workflows** - Users can rerun past tasks, export history sessions as JSON, and approve MCP-triggered tasks before execution
+
+### Improvements
+
+- **Unified versioning across packages** - The extension now follows the root workspace version. Changelog entries are no longer split into a separate extension version section
+- **Configurable `stepDelay`** - Agent pacing between steps is now configurable via `stepDelay`
+- **Optional API key** - `apiKey` can now be omitted for compatible deployments that do not require one
+- **Optional named tool choice** - Tool invocation can disable named tool choice for providers that behave better without it
+- **Better rich-text input support** - Improved `contenteditable` handling with better event dispatching and `execCommand` fallback for more editors
+- **More flexible DOM extraction** - `includeAttributes` now supports wildcards, `contenteditable` is included by default, and heuristically interactive elements expose more useful attributes
+- **MiniMax model support** - Added MiniMax compatibility, with the default recommendation updated to `MiniMax-M2.7`
+
+### Bug Fixes
+
+- Fixed Safari issues when `requestIdleCallback` is unavailable
+- Avoid throwing when `webgl2` initialization fails
+- Improved OpenAI-compatible request patches for GPT-5.4 chat tools and MiniMax temperature/tool-call compatibility
+- Fixed several UI polish issues in the extension and website, including cursor and layout regressions
+
 ## [1.5.1] - 2026-03-05
 
 ### Breaking Changes
@@ -153,42 +178,6 @@ PageAgent is now ready for production use. The API is stable and breaking change
 - **Ask User Tool** - Agent can ask users for clarification
 - **i18n Support** - English and Chinese localization
 
-### Configuration 
-
-```typescript
-// Version 1.0.0
-interface PageAgentConfig {
-    // LLM Configuration (required)
-    baseURL: string
-    apiKey: string
-    model: string
-    temperature?: number
-    maxRetries?: number
-    customFetch?: typeof fetch
-
-    // Agent Configuration
-    language?: 'en-US' | 'zh-CN'
-    maxSteps?: number // default: 20
-    customTools?: Record<string, PageAgentTool> // experimental
-    instructions?: InstructionsConfig
-    transformPageContent?: (content: string) => string | Promise<string>
-    experimentalScriptExecutionTool?: boolean // default: false
-
-    // Lifecycle Hooks (experimental)
-    onBeforeTask?: (agent, result) => void
-    onAfterTask?: (agent, result) => void
-    onBeforeStep?: (agent, stepCount) => void
-    onAfterStep?: (agent, history) => void
-    onDispose?: (agent, reason?) => void
-
-    // Page Controller Configuration
-    enableMask?: boolean // default: true
-    viewportExpansion?: number
-    interactiveBlacklist?: Element[]
-    interactiveWhitelist?: Element[]
-}
-```
-
 ### Packages
 
 | Package                       | Description                        |

docs/README-zh.md

@@ -20,20 +20,20 @@
 ## ✨ Features
 
 - **🎯 轻松集成**
-    - 无需 `浏览器插件` / `Python` / `无头浏览器`。
-    - 纯页面内 JavaScript，一切都在你的网页中完成。
+    - 无需 `浏览器插件` / `Python` / `无头浏览器`，纯页面内 JavaScript
 - **📖 基于文本的 DOM 操作**
-    - 无需截图，无需多模态模型或特殊权限。
-- **🧠 用你自己的 LLM**
-- **🎨 精美 UI，支持人机协同**
-- **🐙 可选的 [Chrome 扩展](https://alibaba.github.io/page-agent/docs/features/chrome-extension)，支持跨页面任务。**
+    - 无需截图，无需多模态模型或特殊权限
+- **🧠 自备 LLM**
+- 🐙 可选的 [Chrome 扩展](https://alibaba.github.io/page-agent/docs/features/chrome-extension)，支持跨页面任务
+    - [MCP Server (Beta)](https://alibaba.github.io/page-agent/docs/features/mcp-server)
 
 ## 💡 应用场景
 
-- **SaaS AI 副驾驶** — 几行代码为你的产品加上 AI 副驾驶，无需重写后端。
+- **SaaS AI Copilot** — 几行代码为你的产品加上 AI 副驾驶，无需重写后端。
 - **智能表单填写** — 把 20 次点击变成一句话。ERP、CRM、管理后台的最佳拍档。
 - **无障碍增强** — 用自然语言让任何网页无障碍。语音指令、屏幕阅读器，零门槛。
-- **跨页面 Agent** — 通过可选的 [Chrome 扩展](https://alibaba.github.io/page-agent/docs/features/chrome-extension)，让你自己的 Agent 跨标签页工作。
+- **跨页面 Agent** — 通过可选的 [Chrome 扩展](https://alibaba.github.io/page-agent/docs/features/chrome-extension)，让你自己的 Web Agent 跨标签页工作。
+- 通过 MCP 为现有 Agent 加入浏览器控制能力。
 
 ## 🚀 快速开始
 
@@ -49,8 +49,8 @@
 
 | Mirrors | URL                                                                                |
 | ------- | ---------------------------------------------------------------------------------- |
-| Global  | https://cdn.jsdelivr.net/npm/page-agent@1.5.9/dist/iife/page-agent.demo.js         |
-| China   | https://registry.npmmirror.com/page-agent/1.5.9/files/dist/iife/page-agent.demo.js |
+| Global  | https://cdn.jsdelivr.net/npm/page-agent@1.7.1/dist/iife/page-agent.demo.js         |
+| China   | https://registry.npmmirror.com/page-agent/1.7.1/files/dist/iife/page-agent.demo.js |
 
 ### NPM 安装
 
@@ -75,11 +75,13 @@ await agent.execute('点击登录按钮')
 
 ## 🤝 贡献
 
-欢迎社区贡献！请参阅 [CONTRIBUTING.md](../CONTRIBUTING.md) 了解安装与贡献指南。请在贡献前阅读[行为准则](CODE_OF_CONDUCT.md)。
+欢迎社区贡献！请参阅 [CONTRIBUTING.md](../CONTRIBUTING.md) 了解安装与贡献指南。
 
-我们不接受未经实质性人类参与、完全由 Bot 或 Agent 自动生成的代码，机器人账号可能被禁止参与互动。
+提交 issue 或 PR 之前，请先阅读[作者声明](https://github.com/alibaba/page-agent/issues/349)和[行为准则](CODE_OF_CONDUCT.md)。
 
-## 👏 致谢
+我们不接受未经实质性人类参与、完全由 Bot 或 Agent 自动生成的代码。
+
+## 👏 声明与致谢
 
 本项目基于 **[`browser-use`](https://github.com/browser-use/browser-use)** 的优秀工作构建。
 
@@ -95,12 +97,9 @@ Licensed under the MIT License
 We gratefully acknowledge the browser-use project and its contributors for their
 excellent work on web automation and DOM interaction patterns that helped make
 this project possible.
-
-Third-party dependencies and their licenses can be found in the package.json
-file and in the node_modules directory after installation.
 ```
 
-## 📄 许可证
+## ⚖️ 许可证
 
 [MIT License](../LICENSE)
 
@@ -108,10 +107,3 @@ file and in the node_modules directory after installation.
 
 **⭐ 如果觉得 PageAgent 有用或有趣，请给项目点个星！**
 
-<a href="https://www.star-history.com/?repos=alibaba%2Fpage-agent&type=date">
- <picture>
-   <source media="(prefers-color-scheme: dark)" srcset="https://api.star-history.com/image?repos=alibaba/page-agent&type=date&theme=dark&legend=top-left&v=7" />
-   <source media="(prefers-color-scheme: light)" srcset="https://api.star-history.com/image?repos=alibaba/page-agent&type=date&legend=top-left&v=7" />
-   <img alt="Star History Chart" src="https://api.star-history.com/image?repos=alibaba/page-agent&type=date&legend=top-left&v=7" />
- </picture>
-</a>

docs/developer-guide.md

@@ -0,0 +1,117 @@
+# Developer Guide
+
+This file is for local development workflows.
+
+For contribution rules and expectations, see [../CONTRIBUTING.md](../CONTRIBUTING.md).
+
+## 🚀 Quick Start
+
+### Development Setup
+
+1. **Prerequisites**
+    - `macOS` / `Linux` / `WSL`
+    - `node.js >= 20` with `npm >= 10`
+    - An editor that supports `ts/eslint/prettier`
+    - Make sure `eslint`, `prettier` and `commitlint` work well. Un-linted code won't pass the CI.
+
+2. **Setup**
+
+    ```bash
+    npm i
+    npm start          # Start demo and documentation site
+    npm run build      # Build libs and website
+    ```
+
+## 📦 Project Structure
+
+This is a **monorepo** with npm workspaces containing **4 main packages**:
+
+- **Page Agent** (`packages/page-agent/`) - Main entry with built-in UI Panel, published as `page-agent` on npm
+- **Core** (`packages/core/`) - Core agent logic without UI (npm: `@page-agent/core`)
+- **Extension** (`packages/extension/`) - Chrome extension for multi-page tasks and browser-level automation
+- **Website** (`packages/website/`) - React documentation and landing page. Also as demo and test page for the core lib. private package `@page-agent/website`
+
+> We use a simplified monorepo solution with `native npm-workspace + ts reference + vite alias`. No fancy tooling. Hoisting is required.
+>
+> - When developing. Use alias so that we don't have to pre-build.
+> - When bundling. Use external and disable ts `paths` alias.
+> - When bundling `IIFE` and `Website`. Bundle everything together.
+
+## 🤖 AGENTS.md Alias
+
+If your AI assistant does not support [AGENTS.md](https://agents.md/). Add an alias for it:
+
+- claude-code (`CLAUDE.md`)
+
+    ```markdown
+    @AGENTS.md
+    ```
+
+- antigravity (`.agent/rules/alias.md`)
+
+    ```markdown
+    ---
+    trigger: always_on
+    ---
+
+    @../../AGENTS.md
+    ```
+
+## 🔧 Development Workflows
+
+### Test With Your Own LLM API
+
+- Create a `.env` file in the repo root with your LLM API config
+
+    ```env
+    LLM_MODEL_NAME=gpt-5.2
+    LLM_API_KEY=your-api-key
+    LLM_BASE_URL=https://api.your-llm-provider.com/v1
+    ```
+
+- **Ollama example** (tested on 0.15 + qwen3:14b, RTX3090 24GB):
+
+    ```env
+    LLM_BASE_URL="http://localhost:11434/v1"
+    LLM_API_KEY="NA"
+    LLM_MODEL_NAME="qwen3:14b"
+    ```
+
+    > @see https://alibaba.github.io/page-agent/docs/features/models#ollama for configuration
+
+- **Restart the dev server** to load new env vars
+- If not provided, the demo will use the free testing proxy by default. By using it, you agree to its [terms](./terms-and-privacy.md).
+
+### Extension Development
+
+```bash
+# make sure you ran `npm run build:libs` first and every time you changed the core libs
+npm run dev -w @page-agent/ext
+npm run zip -w @page-agent/ext
+```
+
+- Update `packages/extension/docs/extension_api.md` for API integration details
+
+### Testing on Other Websites
+
+- Start and serve a local `iife` script
+
+    ```bash
+    npm run dev:demo # Serving IIFE with auto rebuild at http://localhost:5174/page-agent.demo.js
+    ```
+
+- Add a new bookmark
+
+    ```javascript
+    javascript:(function(){var s=document.createElement('script');s.src=`http://localhost:5174/page-agent.demo.js?t=${Math.random()}`;s.onload=()=>console.log(%27PageAgent ready!%27);document.head.appendChild(s);})();
+    ```
+
+- Click the bookmark on any page to load Page-Agent
+
+> Warning: AK in your local `.env` will be inlined in the iife script. Be very careful when you distribute the script.
+
+### Adding Documentation
+
+Ask an AI to help you add documentation to the `website/` package. Follow the existing style.
+
+> Our AGENTS.md file and guardrails are designed for this purpose. But please be careful and review anything AI generated.

package.json

@@ -1,7 +1,7 @@
 {
     "name": "root",
     "private": true,
-    "version": "1.5.9",
+    "version": "1.7.1",
     "type": "module",
     "workspaces": [
         "packages/page-controller",
@@ -42,7 +42,7 @@
         "@commitlint/config-conventional": "^20.5.0",
         "@eslint/js": "^9.39.2",
         "@microsoft/api-extractor": "^7.57.7",
-        "@tailwindcss/vite": "^4.2.1",
+        "@tailwindcss/vite": "^4.2.2",
         "@trivago/prettier-plugin-sort-imports": "^6.0.2",
         "@types/node": "^25.5.0",
         "@vitejs/plugin-react-swc": "^4.3.0",
@@ -60,7 +60,7 @@
         "lint-staged": "^16.4.0",
         "prettier": "^3.8.0",
         "typescript": "^5.9.3",
-        "typescript-eslint": "^8.57.1",
+        "typescript-eslint": "^8.58.0",
         "unplugin-dts": "^1.0.0-beta.6",
         "vite": "^7.3.1",
         "vite-plugin-css-injected-by-js": "^4.0.1",

packages/core/package.json

@@ -1,7 +1,7 @@
 {
     "name": "@page-agent/core",
     "private": false,
-    "version": "1.5.9",
+    "version": "1.7.1",
     "type": "module",
     "main": "./dist/esm/page-agent-core.js",
     "module": "./dist/esm/page-agent-core.js",
@@ -44,8 +44,8 @@
     },
     "dependencies": {
         "chalk": "^5.6.2",
-        "@page-agent/llms": "1.5.9",
-        "@page-agent/page-controller": "1.5.9"
+        "@page-agent/llms": "1.7.1",
+        "@page-agent/page-controller": "1.7.1"
     },
     "peerDependencies": {
         "zod": "^3.25.0 || ^4.0.0"

packages/core/src/tools/index.ts

@@ -131,7 +131,8 @@ tools.set(
 tools.set(
 	'scroll',
 	tool({
-		description: 'Scroll the page vertically. Use index for scroll elements (dropdowns/custom UI).',
+		description:
+			'Scroll vertically. Without index: scrolls the document. With index: scrolls the container at that index (or its nearest scrollable ancestor). Use index of a data-scrollable element to scroll a specific area.',
 		inputSchema: z.object({
 			down: z.boolean().default(true),
 			num_pages: z.number().min(0).max(10).optional().default(0.1),
@@ -155,7 +156,7 @@ tools.set(
 	'scroll_horizontally',
 	tool({
 		description:
-			'Scroll the page horizontally, or within a specific element by index. Useful for wide tables.',
+			'Scroll horizontally. Without index: scrolls the document. With index: scrolls the container at that index (or its nearest scrollable ancestor). Use index of a data-scrollable element to scroll a specific area.',
 		inputSchema: z.object({
 			right: z.boolean().default(true),
 			pixels: z.number().int().min(0),

packages/extension/docs/extension_api.md

@@ -115,12 +115,21 @@ import type {
 
 export interface ExecuteConfig {
   baseURL: string
-  apiKey: string
   model: string
+  apiKey?: string
+
+  // Global system-level instructions for the agent.
+  // Equivalent to AgentConfig.instructions.system.
+  systemInstruction?: string
 
   // Include the initial tab where page JS starts. Default: true.
   includeInitialTab?: boolean
 
+  // Control all unpinned tabs in the window instead of only the tab group.
+  // When enabled, agent sees and can switch to every non-pinned tab.
+  // Default: false. Experimental.
+  experimentalIncludeAllTabs?: boolean
+
   onStatusChange?: (status: AgentStatus) => void
   onActivity?: (activity: AgentActivity) => void
   onHistoryUpdate?: (history: HistoricalEvent[]) => void
@@ -205,9 +214,13 @@ import type {
 
 interface ExecuteConfig {
   baseURL: string
-  apiKey: string
   model: string
+  apiKey?: string
+
+  systemInstruction?: string
+
   includeInitialTab?: boolean
+  experimentalIncludeAllTabs?: boolean
   onStatusChange?: (status: AgentStatus) => void
   onActivity?: (activity: AgentActivity) => void
   onHistoryUpdate?: (history: HistoricalEvent[]) => void

packages/extension/package.json

@@ -1,7 +1,7 @@
 {
     "name": "@page-agent/ext",
     "private": true,
-    "version": "1.5.9",
+    "version": "1.7.1",
     "type": "module",
     "scripts": {
         "dev": "wxt",
@@ -16,32 +16,31 @@
         "@radix-ui/react-separator": "^1.1.8",
         "@radix-ui/react-slot": "^1.2.4",
         "@radix-ui/react-switch": "^1.2.6",
-        "@radix-ui/react-tooltip": "^1.2.8",
-        "@types/chrome": "^0.1.37",
+        "@types/chrome": "^0.1.38",
         "@types/react": "^19.2.14",
         "@types/react-dom": "^19.2.1",
         "@wxt-dev/module-react": "^1.2.2",
         "class-variance-authority": "^0.7.1",
         "clsx": "^2.1.1",
         "idb": "^8.0.3",
-        "lucide-react": "^0.577.0",
-        "motion": "^12.37.0",
+        "lucide-react": "^1.7.0",
+        "motion": "^12.38.0",
         "next-themes": "^0.4.6",
         "react": "^19.2.4",
         "react-dom": "^19.2.4",
         "rough-notation": "^0.5.1",
-        "simple-icons": "^16.12.0",
+        "simple-icons": "^16.14.0",
         "sonner": "^2.0.7",
         "tailwind-merge": "^3.5.0",
         "tailwindcss": "^4.1.14",
         "tw-animate-css": "^1.4.0",
-        "wxt": "^0.20.19"
+        "wxt": "^0.20.20"
     },
     "dependencies": {
-        "@page-agent/core": "1.5.9",
-        "@page-agent/llms": "1.5.9",
-        "@page-agent/page-controller": "1.5.9",
-        "@page-agent/ui": "1.5.9",
+        "@page-agent/core": "1.7.1",
+        "@page-agent/llms": "1.7.1",
+        "@page-agent/page-controller": "1.7.1",
+        "@page-agent/ui": "1.7.1",
         "ai-motion": "^0.4.8",
         "chalk": "^5.6.2"
     },

packages/extension/src/agent/MultiPageAgent.ts

@@ -11,13 +11,18 @@ function detectLanguage(): 'en-US' | 'zh-CN' {
 	return lang.startsWith('zh') ? 'zh-CN' : 'en-US'
 }
 
+interface MultiPageAgentConfig extends AgentConfig {
+	includeInitialTab?: boolean
+	experimentalIncludeAllTabs?: boolean
+}
+
 /**
  * MultiPageAgent
  * - use with extension
  * - can be used from a side panel or a content script
  */
 export class MultiPageAgent extends PageAgentCore {
-	constructor(config: AgentConfig & { includeInitialTab?: boolean }) {
+	constructor(config: MultiPageAgentConfig) {
 		// multi page controller
 		const tabsController = new TabsController()
 		const pageController = new RemotePageController(tabsController)
@@ -31,8 +36,8 @@ export class MultiPageAgent extends PageAgentCore {
 			`Default working language: **${targetLanguage}**`
 		)
 
-		// include initial tab for controlling
 		const includeInitialTab = config.includeInitialTab ?? true
+		const experimentalIncludeAllTabs = config.experimentalIncludeAllTabs ?? false
 
 		/**
 		 * When the agent is in side-panel and user closed the side-panel.
@@ -50,7 +55,7 @@ export class MultiPageAgent extends PageAgentCore {
 			customSystemPrompt: systemPrompt,
 
 			onBeforeTask: async (agent) => {
-				await tabsController.init(agent.task, includeInitialTab)
+				await tabsController.init(agent.task, { includeInitialTab, experimentalIncludeAllTabs })
 
 				heartBeatInterval = window.setInterval(() => {
 					chrome.storage.local.set({
@@ -75,6 +80,7 @@ export class MultiPageAgent extends PageAgentCore {
 			},
 
 			onBeforeStep: async (agent) => {
+				if (!tabsController.currentTabId) return
 				// make sure the current tab is loaded before the step starts
 				await tabsController.waitUntilTabLoaded(tabsController.currentTabId!)
 			},

packages/extension/src/agent/RemotePageController.background.ts

@@ -10,9 +10,7 @@ export function handlePageControlMessage(
 ): true | undefined {
 	const PREFIX = '[RemotePageController.background]'
 
-	function debug(...messages: any[]) {
-		console.debug(`\x1b[90m${PREFIX}\x1b[0m`, ...messages)
-	}
+	const debug = console.debug.bind(console, `\x1b[90m${PREFIX}\x1b[0m`)
 
 	const { action, payload, targetTabId } = message
 

packages/extension/src/agent/RemotePageController.content.ts

@@ -19,7 +19,10 @@ export function initPageController() {
 
 	function getPC(): PageController {
 		if (!pageController) {
-			pageController = new PageController({ enableMask: false, viewportExpansion: 400 })
+			pageController = new PageController({
+				enableMask: false,
+				viewportExpansion: 400,
+			})
 		}
 		return pageController
 	}

packages/extension/src/agent/RemotePageController.ts

@@ -4,9 +4,7 @@ import type { TabsController } from './TabsController'
 
 const PREFIX = '[RemotePageController]'
 
-function debug(...messages: any[]) {
-	console.debug(`\x1b[90m${PREFIX}\x1b[0m`, ...messages)
-}
+const debug = console.debug.bind(console, `\x1b[90m${PREFIX}\x1b[0m`)
 
 function sendMessage(message: {
 	type: 'PAGE_CONTROL'
@@ -58,8 +56,6 @@ export class RemotePageController {
 	}
 
 	async getBrowserState(): Promise<BrowserState> {
-		if (!this.currentTabId) throw new Error('tabsController not initialized.')
-
 		let browserState = {} as BrowserState
 		debug('getBrowserState', this.currentTabId)
 
@@ -178,7 +174,7 @@ interface DomActionReturn {
 /**
  * Check if a URL can run content scripts.
  */
-function isContentScriptAllowed(url: string | undefined): boolean {
+export function isContentScriptAllowed(url: string | undefined): boolean {
 	if (!url) return false
 
 	const restrictedPatterns = [

packages/extension/src/agent/TabsController.background.ts

@@ -5,9 +5,7 @@ import type { TabAction } from './TabsController'
 
 const PREFIX = '[TabsController.background]'
 
-function debug(...messages: any[]) {
-	console.debug(`\x1b[90m${PREFIX}\x1b[0m`, ...messages)
-}
+const debug = console.debug.bind(console, `\x1b[90m${PREFIX}\x1b[0m`)
 
 export function handleTabControlMessage(
 	message: { type: 'TAB_CONTROL'; action: TabAction; payload: any },
@@ -20,11 +18,10 @@ export function handleTabControlMessage(
 		case 'get_active_tab': {
 			debug('get_active_tab')
 			chrome.tabs
-				.query({ active: true, currentWindow: true })
+				.query({ active: true })
 				.then((tabs) => {
-					const tabId = tabs.length > 0 ? tabs[0].id || null : null
-					debug('get_active_tab: success', tabId)
-					sendResponse({ success: true, tabId })
+					debug('get_active_tab: success', tabs)
+					sendResponse({ success: true, tab: tabs[0] })
 				})
 				.catch((error) => {
 					sendResponse({ error: error instanceof Error ? error.message : String(error) })
@@ -63,7 +60,7 @@ export function handleTabControlMessage(
 		case 'create_tab_group': {
 			debug('create_tab_group', payload)
 			chrome.tabs
-				.group({ tabIds: payload.tabIds })
+				.group({ tabIds: payload.tabIds, createProperties: { windowId: payload.windowId } })
 				.then((groupId) => {
 					debug('create_tab_group: success', groupId)
 					sendResponse({ success: true, groupId })
@@ -114,47 +111,59 @@ export function handleTabControlMessage(
 			return true // async response
 		}
 
+		case 'get_window_tabs': {
+			debug('get_window_tabs', payload)
+			chrome.tabs
+				.query({ windowId: payload.windowId })
+				.then((tabs) => {
+					sendResponse({ success: true, tabs })
+				})
+				.catch((error) => {
+					sendResponse({ error: error instanceof Error ? error.message : String(error) })
+				})
+			return true
+		}
+
 		default:
 			sendResponse({ error: `Unknown action: ${action}` })
 			return
 	}
 }
 
-export function setupTabChangeEvents() {
-	console.log('[TabsController.background] setupTabChangeEvents')
+const tabEventPorts = new Set<chrome.runtime.Port>()
+
+function broadcastTabEvent(message: object) {
+	for (const port of tabEventPorts) {
+		port.postMessage(message)
+	}
+}
+
+/**
+ * Port-based tab events: agents connect via `chrome.runtime.connect({ name: 'tab-events' })`
+ * and receive tab change events through the port. Works for both extension pages and content scripts.
+ */
+export function setupTabEventsPort() {
+	chrome.runtime.onConnect.addListener((port) => {
+		if (port.name !== 'tab-events') return
+
+		debug('port connected', port.sender?.tab?.id ?? port.sender?.url)
+		tabEventPorts.add(port)
+
+		port.onDisconnect.addListener(() => {
+			debug('port disconnected')
+			tabEventPorts.delete(port)
+		})
+	})
 
 	chrome.tabs.onCreated.addListener((tab) => {
-		debug('onCreated', tab)
-		chrome.runtime
-			.sendMessage({ type: 'TAB_CHANGE', action: 'created', payload: { tab } })
-			.catch((error) => {
-				debug('onCreated error:', error)
-			})
+		broadcastTabEvent({ action: 'created', payload: { tab } })
 	})
 
 	chrome.tabs.onRemoved.addListener((tabId, removeInfo) => {
-		debug('onRemoved', tabId, removeInfo)
-		chrome.runtime
-			.sendMessage({
-				type: 'TAB_CHANGE',
-				action: 'removed',
-				payload: { tabId, removeInfo },
-			})
-			.catch((error) => {
-				debug('onRemoved error:', error)
-			})
+		broadcastTabEvent({ action: 'removed', payload: { tabId, removeInfo } })
 	})
 
 	chrome.tabs.onUpdated.addListener((tabId, changeInfo, tab) => {
-		debug('onUpdated', tabId, changeInfo)
-		chrome.runtime
-			.sendMessage({
-				type: 'TAB_CHANGE',
-				action: 'updated',
-				payload: { tabId, changeInfo, tab },
-			})
-			.catch((error) => {
-				debug('onUpdated error:', error)
-			})
+		broadcastTabEvent({ action: 'updated', payload: { tabId, changeInfo, tab } })
 	})
 }

packages/extension/src/agent/TabsController.ts

@@ -1,8 +1,8 @@
+import { isContentScriptAllowed } from './RemotePageController'
+
 const PREFIX = '[TabsController]'
 
-function debug(...messages: any[]) {
-	console.debug(`\x1b[90m${PREFIX}\x1b[0m`, ...messages)
-}
+const debug = console.debug.bind(console, `\x1b[90m${PREFIX}\x1b[0m`)
 
 function sendMessage(message: {
 	type: 'TAB_CONTROL'
@@ -20,101 +20,102 @@ function sendMessage(message: {
  * - live in the agent env (extension page or content script)
  * - no chrome apis. call sw for tab operations
  */
-export class TabsController extends EventTarget {
+export class TabsController {
 	currentTabId: number | null = null
 
+	private disposed = false
+	private port: chrome.runtime.Port | null = null
+	private portRetries = 0
+
+	private windowId: number | null = null
 	private tabs: TabMeta[] = []
 	private initialTabId: number | null = null
 	private tabGroupId: number | null = null
+	private experimentalIncludeAllTabs = false
 	private task: string = ''
 
-	async init(task: string, includeInitialTab: boolean = true) {
-		debug('init', task, includeInitialTab)
+	async init(task: string, options: TabsInitOptions = {}) {
+		const { includeInitialTab = true, experimentalIncludeAllTabs = false } = options
+		debug('init', task, options)
+
+		if (this.disposed) {
+			throw new Error('TabsController already disposed')
+		}
 
-		this.task = task
-		this.tabs = []
 		this.currentTabId = null
+		this.disposed = false
+		this.port = null
+		this.portRetries = 0
+
+		this.windowId = null
+		this.tabs = []
 		this.tabGroupId = null
 		this.initialTabId = null
+		this.experimentalIncludeAllTabs = experimentalIncludeAllTabs
+		this.task = task
 
-		const result = await sendMessage({
+		const activeTabResult = await sendMessage({
 			type: 'TAB_CONTROL',
 			action: 'get_active_tab',
 		})
 
-		this.initialTabId = result.tabId
+		this.initialTabId = activeTabResult.tab?.id
+		this.windowId = activeTabResult.tab?.windowId
 
-		if (!this.initialTabId) {
-			throw new Error('Failed to get initial tab ID')
+		if (!this.initialTabId || !this.windowId) {
+			if (activeTabResult.error) {
+				throw new Error(activeTabResult.error)
+			} else {
+				throw new Error('Failed to get active tab')
+			}
 		}
 
-		if (includeInitialTab) {
-			this.currentTabId = this.initialTabId
+		this.connectTabEvents()
 
+		if (experimentalIncludeAllTabs) {
+			const allTabs = await sendMessage({
+				type: 'TAB_CONTROL',
+				action: 'get_window_tabs',
+				payload: { windowId: this.windowId },
+			})
+			for (const tab of allTabs.tabs as chrome.tabs.Tab[]) {
+				if (tab.id && !tab.pinned && isContentScriptAllowed(tab.url)) {
+					this.addTab({
+						id: tab.id,
+						isInitial: tab.id === this.initialTabId,
+						url: tab.url,
+						title: tab.title,
+						status: tab.status,
+					})
+				}
+			}
+			if (this.tabs.find((t) => t.id === this.initialTabId)) {
+				this.currentTabId = this.initialTabId
+				await this.createTabGroup([this.initialTabId])
+			}
+		} else if (includeInitialTab) {
 			const info = await sendMessage({
 				type: 'TAB_CONTROL',
 				action: 'get_tab_info',
 				payload: { tabId: this.initialTabId },
 			})
 
-			this.tabs.push({
-				id: result.tabId,
-				isInitial: true,
-				url: info.url,
-				title: info.title,
-				status: info.status,
-			})
+			if (isContentScriptAllowed(info.url) && !info.pinned) {
+				this.currentTabId = this.initialTabId
 
-			await this.createTabGroup([this.initialTabId])
+				this.addTab({
+					id: this.initialTabId,
+					isInitial: true,
+					url: info.url,
+					title: info.title,
+					status: info.status,
+				})
+
+				await this.createTabGroup([this.initialTabId])
+			}
 		}
 
 		await this.updateCurrentTabId(this.currentTabId)
-
-		const tabChangeHandler = (message: any): void => {
-			if (message.type !== 'TAB_CHANGE') {
-				// throw new Error(`[TabsController]: Invalid message type: ${message.type}`)
-				return
-			}
-
-			if (message.action === 'created') {
-				const tab = message.payload.tab as chrome.tabs.Tab
-				if (tab.groupId === this.tabGroupId && tab.id != null) {
-					// Tab created in our controlled group
-					if (!this.tabs.find((t) => t.id === tab.id)) {
-						this.tabs.push({ id: tab.id, isInitial: false })
-					}
-					this.switchToTab(tab.id)
-				}
-			} else if (message.action === 'removed') {
-				const { tabId } = message.payload as { tabId: number }
-				const targetTab = this.tabs.find((t) => t.id === tabId)
-				if (targetTab) {
-					this.tabs = this.tabs.filter((t) => t.id !== tabId)
-					if (this.currentTabId === tabId) {
-						const newCurrentTab = this.tabs[this.tabs.length - 1] || null
-						if (newCurrentTab) {
-							this.switchToTab(newCurrentTab.id)
-						} else {
-							this.updateCurrentTabId(null)
-						}
-					}
-				}
-			} else if (message.action === 'updated') {
-				const { tabId, tab } = message.payload as { tabId: number; tab: chrome.tabs.Tab }
-				const targetTab = this.tabs.find((t) => t.id === tabId)
-				if (targetTab) {
-					targetTab.url = tab.url
-					targetTab.title = tab.title
-					targetTab.status = tab.status
-				}
-			}
-		}
-
-		chrome.runtime.onMessage.addListener(tabChangeHandler)
-
-		this.addEventListener('dispose', () => {
-			chrome.runtime.onMessage.removeListener(tabChangeHandler)
-		})
 	}
 
 	async openNewTab(url: string): Promise<string> {
@@ -132,7 +133,7 @@ export class TabsController extends EventTarget {
 
 		const tabId = result.tabId as number
 
-		this.tabs.push({
+		this.addTab({
 			id: tabId,
 			isInitial: false,
 		})
@@ -205,7 +206,7 @@ export class TabsController extends EventTarget {
 		const result = await sendMessage({
 			type: 'TAB_CONTROL',
 			action: 'create_tab_group',
-			payload: { tabIds },
+			payload: { tabIds, windowId: this.windowId },
 		})
 
 		if (!result?.success) {
@@ -228,6 +229,11 @@ export class TabsController extends EventTarget {
 		})
 	}
 
+	private addTab(meta: TabMeta) {
+		if (this.tabs.find((t) => t.id === meta.id)) return
+		this.tabs.push(meta)
+	}
+
 	async updateCurrentTabId(tabId: number | null) {
 		debug('updateCurrentTabId', tabId)
 
@@ -284,9 +290,77 @@ export class TabsController extends EventTarget {
 		await waitUntil(() => tab.status === 'complete', 4_000)
 	}
 
-	dispose() {
-		this.dispatchEvent(new Event('dispose'))
+	/**
+	 * Connect to background SW via port to receive tab change events.
+	 *
+	 * @note Port is 1:1 (runtime.connect → background SW has no frames),
+	 * so onDisconnect fires exactly once and we can safely reconnect.
+	 * Reconnection may miss events during the gap.
+	 * TODO: refresh this.tabs from background after reconnect to stay consistent.
+	 */
+	private connectTabEvents() {
+		this.port = chrome.runtime.connect({ name: 'tab-events' })
+
+		this.port.onMessage.addListener((message: any) => {
+			if (this.disposed) return
+			this.portRetries = 0
+
+			if (message.action === 'created') {
+				const tab = message.payload.tab as chrome.tabs.Tab
+				const shouldTrack = this.experimentalIncludeAllTabs || tab.groupId === this.tabGroupId
+				if (shouldTrack && tab.id != null) {
+					this.addTab({ id: tab.id, isInitial: false })
+					this.switchToTab(tab.id)
+				}
+			} else if (message.action === 'removed') {
+				const { tabId } = message.payload as { tabId: number }
+				const targetTab = this.tabs.find((t) => t.id === tabId)
+				if (targetTab) {
+					this.tabs = this.tabs.filter((t) => t.id !== tabId)
+					if (this.currentTabId === tabId) {
+						const newCurrentTab = this.tabs[this.tabs.length - 1] || null
+						if (newCurrentTab) {
+							this.switchToTab(newCurrentTab.id)
+						} else {
+							this.updateCurrentTabId(null)
+						}
+					}
+				}
+			} else if (message.action === 'updated') {
+				const { tabId, tab } = message.payload as { tabId: number; tab: chrome.tabs.Tab }
+				const targetTab = this.tabs.find((t) => t.id === tabId)
+				if (targetTab) {
+					targetTab.url = tab.url
+					targetTab.title = tab.title
+					targetTab.status = tab.status
+				}
+			}
+		})
+
+		this.port.onDisconnect.addListener(() => {
+			this.port = null
+			if (this.disposed) return
+			if (this.portRetries >= 7) {
+				console.error(PREFIX, 'tab events port failed after 7 retries, giving up')
+				return
+			}
+			debug('port disconnected, reconnecting...')
+			this.portRetries++
+			this.connectTabEvents()
+		})
 	}
+
+	dispose() {
+		debug('dispose')
+		this.disposed = true
+		this.port?.disconnect()
+		this.port = null
+	}
+}
+
+export interface TabsInitOptions {
+	includeInitialTab?: boolean
+	experimentalIncludeAllTabs?: boolean
 }
 
 export type TabAction =
@@ -298,6 +372,7 @@ export type TabAction =
 	| 'add_tab_to_group'
 	| 'close_tab'
 	| 'get_tab_title'
+	| 'get_window_tabs'
 
 interface TabMeta {
 	id: number

packages/extension/src/agent/constants.ts

@@ -3,12 +3,12 @@ import type { LLMConfig } from '@page-agent/llms'
 // Demo LLM for testing
 export const DEMO_MODEL = 'qwen3.5-plus'
 export const DEMO_BASE_URL = 'https://page-ag-testing-ohftxirgbn.cn-shanghai.fcapp.run'
-export const DEMO_API_KEY = 'NA'
+// export const DEMO_API_KEY = 'NA'
 
 export const DEMO_CONFIG: LLMConfig = {
-	apiKey: DEMO_API_KEY,
 	baseURL: DEMO_BASE_URL,
 	model: DEMO_MODEL,
+	// apiKey: DEMO_API_KEY,
 }
 
 /** Legacy testing endpoints that should be auto-migrated to DEMO_BASE_URL */

packages/extension/src/agent/useAgent.ts

@@ -21,6 +21,8 @@ export interface AdvancedConfig {
 	maxSteps?: number
 	systemInstruction?: string
 	experimentalLlmsTxt?: boolean
+	experimentalIncludeAllTabs?: boolean
+	disableNamedToolChoice?: boolean
 }
 
 export interface ExtConfig extends LLMConfig, AdvancedConfig {
@@ -124,6 +126,8 @@ export function useAgent(): UseAgentResult {
 			maxSteps,
 			systemInstruction,
 			experimentalLlmsTxt,
+			experimentalIncludeAllTabs,
+			disableNamedToolChoice,
 			...llmConfig
 		}: ExtConfig) => {
 			await chrome.storage.local.set({ llmConfig })
@@ -132,7 +136,13 @@ export function useAgent(): UseAgentResult {
 			} else {
 				await chrome.storage.local.remove('language')
 			}
-			const advancedConfig: AdvancedConfig = { maxSteps, systemInstruction, experimentalLlmsTxt }
+			const advancedConfig: AdvancedConfig = {
+				maxSteps,
+				systemInstruction,
+				experimentalLlmsTxt,
+				experimentalIncludeAllTabs,
+				disableNamedToolChoice,
+			}
 			await chrome.storage.local.set({ advancedConfig })
 			setConfig({ ...llmConfig, ...advancedConfig, language })
 		},

packages/extension/src/components/ConfigPanel.tsx

@@ -1,18 +1,20 @@
 import {
-	ChevronDown,
 	Copy,
 	CornerUpLeft,
+	ExternalLink,
 	Eye,
 	EyeOff,
+	FoldVertical,
 	HatGlasses,
 	Home,
 	Loader2,
 	Scale,
+	UnfoldVertical,
 } from 'lucide-react'
 import { useEffect, useState } from 'react'
 import { siGithub } from 'simple-icons'
 
-import { DEMO_API_KEY, DEMO_BASE_URL, DEMO_MODEL, isTestingEndpoint } from '@/agent/constants'
+import { DEMO_BASE_URL, DEMO_MODEL, isTestingEndpoint } from '@/agent/constants'
 import type { ExtConfig, LanguagePreference } from '@/agent/useAgent'
 import { Button } from '@/components/ui/button'
 import { Input } from '@/components/ui/input'
@@ -25,30 +27,38 @@ interface ConfigPanelProps {
 }
 
 export function ConfigPanel({ config, onSave, onClose }: ConfigPanelProps) {
-	const [apiKey, setApiKey] = useState(config?.apiKey || DEMO_API_KEY)
 	const [baseURL, setBaseURL] = useState(config?.baseURL || DEMO_BASE_URL)
 	const [model, setModel] = useState(config?.model || DEMO_MODEL)
+	const [apiKey, setApiKey] = useState(config?.apiKey)
 	const [language, setLanguage] = useState<LanguagePreference>(config?.language)
-	const [maxSteps, setMaxSteps] = useState<number | undefined>(config?.maxSteps)
+	const [maxSteps, setMaxSteps] = useState(config?.maxSteps)
 	const [systemInstruction, setSystemInstruction] = useState(config?.systemInstruction ?? '')
 	const [experimentalLlmsTxt, setExperimentalLlmsTxt] = useState(
 		config?.experimentalLlmsTxt ?? false
 	)
+	const [experimentalIncludeAllTabs, setExperimentalIncludeAllTabs] = useState(
+		config?.experimentalIncludeAllTabs ?? false
+	)
+	const [disableNamedToolChoice, setDisableNamedToolChoice] = useState(
+		config?.disableNamedToolChoice ?? false
+	)
 	const [advancedOpen, setAdvancedOpen] = useState(false)
 	const [saving, setSaving] = useState(false)
-	const [userAuthToken, setUserAuthToken] = useState<string>('')
+	const [userAuthToken, setUserAuthToken] = useState('')
 	const [copied, setCopied] = useState(false)
 	const [showToken, setShowToken] = useState(false)
 	const [showApiKey, setShowApiKey] = useState(false)
 
 	useEffect(() => {
-		setApiKey(config?.apiKey || DEMO_API_KEY)
 		setBaseURL(config?.baseURL || DEMO_BASE_URL)
 		setModel(config?.model || DEMO_MODEL)
+		setApiKey(config?.apiKey)
 		setLanguage(config?.language)
 		setMaxSteps(config?.maxSteps)
 		setSystemInstruction(config?.systemInstruction ?? '')
 		setExperimentalLlmsTxt(config?.experimentalLlmsTxt ?? false)
+		setExperimentalIncludeAllTabs(config?.experimentalIncludeAllTabs ?? false)
+		setDisableNamedToolChoice(config?.disableNamedToolChoice ?? false)
 	}, [config])
 
 	// Poll for user auth token every second until found
@@ -94,6 +104,8 @@ export function ConfigPanel({ config, onSave, onClose }: ConfigPanelProps) {
 				maxSteps: maxSteps || undefined,
 				systemInstruction: systemInstruction || undefined,
 				experimentalLlmsTxt,
+				experimentalIncludeAllTabs,
+				disableNamedToolChoice,
 			})
 		} finally {
 			setSaving(false)
@@ -153,6 +165,16 @@ export function ConfigPanel({ config, onSave, onClose }: ConfigPanelProps) {
 				</div>
 			</div>
 
+			{/* Hub link */}
+			<a
+				href="/hub.html"
+				target="_blank"
+				className="flex items-center justify-between p-3 rounded-md border bg-muted/50 text-xs font-medium text-muted-foreground hover:text-foreground hover:border-foreground/20 transition-colors"
+			>
+				Manage Page Agent Hub
+				<ExternalLink className="size-3" />
+			</a>
+
 			<div className="flex flex-col gap-1.5">
 				<label className="text-xs text-muted-foreground">Base URL</label>
 				<Input
@@ -167,7 +189,7 @@ export function ConfigPanel({ config, onSave, onClose }: ConfigPanelProps) {
 			{isTestingEndpoint(baseURL) && (
 				<div className="p-2.5 rounded-md border border-amber-500/30 bg-amber-500/5 text-[11px] text-muted-foreground leading-relaxed">
 					<Scale className="size-3 inline-block mr-1 -mt-0.5 text-amber-600" />
-					You are using the free testing API. By using this service you agree to the{' '}
+					You are using our testing API. By using this you agree to the{' '}
 					<a
 						href="https://github.com/alibaba/page-agent/blob/main/docs/terms-and-privacy.md"
 						target="_blank"
@@ -176,14 +198,13 @@ export function ConfigPanel({ config, onSave, onClose }: ConfigPanelProps) {
 					>
 						Terms of Use & Privacy Policy
 					</a>
-					. No sensitive data. No guaranteed availability.
 				</div>
 			)}
 
 			<div className="flex flex-col gap-1.5">
 				<label className="text-xs text-muted-foreground">Model</label>
 				<Input
-					placeholder="gpt-5.2"
+					placeholder="gpt-5.1"
 					value={model}
 					onChange={(e) => setModel(e.target.value)}
 					className="text-xs h-8"
@@ -195,7 +216,7 @@ export function ConfigPanel({ config, onSave, onClose }: ConfigPanelProps) {
 				<div className="flex gap-2 items-center">
 					<Input
 						type={showApiKey ? 'text' : 'password'}
-						placeholder="sk-..."
+						// placeholder="sk-..."
 						value={apiKey}
 						onChange={(e) => setApiKey(e.target.value)}
 						className="text-xs h-8"
@@ -212,7 +233,7 @@ export function ConfigPanel({ config, onSave, onClose }: ConfigPanelProps) {
 			</div>
 
 			<div className="flex flex-col gap-1.5">
-				<label className="text-xs text-muted-foreground">Language</label>
+				<label className="text-xs text-muted-foreground">Response Language</label>
 				<select
 					value={language ?? ''}
 					onChange={(e) => setLanguage((e.target.value || undefined) as LanguagePreference)}
@@ -231,10 +252,7 @@ export function ConfigPanel({ config, onSave, onClose }: ConfigPanelProps) {
 				className="flex items-center gap-1 text-xs text-muted-foreground hover:text-foreground cursor-pointer mt-1 font-bold"
 			>
 				Advanced
-				<ChevronDown
-					className="size-3 transition-transform"
-					style={{ transform: advancedOpen ? 'rotate(0deg)' : 'rotate(90deg)' }}
-				/>
+				{advancedOpen ? <FoldVertical className="size-3" /> : <UnfoldVertical className="size-3" />}
 			</button>
 
 			{advancedOpen && (
@@ -263,10 +281,23 @@ export function ConfigPanel({ config, onSave, onClose }: ConfigPanelProps) {
 						/>
 					</div>
 
+					<label className="flex items-center justify-between cursor-pointer">
+						<span className="text-xs text-muted-foreground">Disable named tool_choice</span>
+						<Switch checked={disableNamedToolChoice} onCheckedChange={setDisableNamedToolChoice} />
+					</label>
+
 					<label className="flex items-center justify-between cursor-pointer">
 						<span className="text-xs text-muted-foreground">Experimental llms.txt support</span>
 						<Switch checked={experimentalLlmsTxt} onCheckedChange={setExperimentalLlmsTxt} />
 					</label>
+
+					<label className="flex items-center justify-between cursor-pointer">
+						<span className="text-xs text-muted-foreground">Experimental include all tabs</span>
+						<Switch
+							checked={experimentalIncludeAllTabs}
+							onCheckedChange={setExperimentalIncludeAllTabs}
+						/>
+					</label>
 				</>
 			)}
 

packages/extension/src/components/HistoryDetail.tsx

@@ -1,12 +1,20 @@
-import { ArrowLeft } from 'lucide-react'
+import { ArrowLeft, RotateCcw, Trash2 } from 'lucide-react'
 import { useEffect, useState } from 'react'
 
 import { Button } from '@/components/ui/button'
-import { type SessionRecord, getSession } from '@/lib/db'
+import { type SessionRecord, deleteSession, getSession } from '@/lib/db'
 
 import { EventCard } from './cards'
 
-export function HistoryDetail({ sessionId, onBack }: { sessionId: string; onBack: () => void }) {
+export function HistoryDetail({
+	sessionId,
+	onBack,
+	onRerun,
+}: {
+	sessionId: string
+	onBack: () => void
+	onRerun: (task: string) => void
+}) {
 	const [session, setSession] = useState<SessionRecord | null>(null)
 
 	useEffect(() => {
@@ -37,6 +45,27 @@ export function HistoryDetail({ sessionId, onBack }: { sessionId: string; onBack
 				<div className="text-xs font-medium" title={session.task}>
 					{session.task}
 				</div>
+				<div className="mt-2 flex items-center gap-2">
+					<button
+						type="button"
+						onClick={() => onRerun(session.task)}
+						className="flex items-center gap-1 text-[10px] text-muted-foreground hover:text-foreground transition-colors cursor-pointer disabled:opacity-50 disabled:cursor-not-allowed"
+					>
+						<RotateCcw className="size-3" />
+						Run again
+					</button>
+					<button
+						type="button"
+						onClick={async () => {
+							await deleteSession(sessionId)
+							onBack()
+						}}
+						className="flex items-center gap-1 text-[10px] text-muted-foreground hover:text-destructive transition-colors cursor-pointer"
+					>
+						<Trash2 className="size-3" />
+						Delete
+					</button>
+				</div>
 			</div>
 
 			{/* Events (read-only) */}

packages/extension/src/components/HistoryList.tsx

@@ -1,8 +1,9 @@
-import { ArrowLeft, CheckCircle, Trash2, XCircle } from 'lucide-react'
+import { ArrowDownToLine, ArrowLeft, CheckCircle, RotateCcw, Trash2, XCircle } from 'lucide-react'
 import { useCallback, useEffect, useState } from 'react'
 
 import { Button } from '@/components/ui/button'
 import { type SessionRecord, clearSessions, deleteSession, listSessions } from '@/lib/db'
+import { downloadHistoryExport } from '@/lib/history-export'
 
 function timeAgo(ts: number): string {
 	const seconds = Math.floor((Date.now() - ts) / 1000)
@@ -18,9 +19,11 @@ function timeAgo(ts: number): string {
 export function HistoryList({
 	onSelect,
 	onBack,
+	onRerun,
 }: {
 	onSelect: (id: string) => void
 	onBack: () => void
+	onRerun: (task: string) => void
 }) {
 	const [sessions, setSessions] = useState<SessionRecord[]>([])
 	const [loading, setLoading] = useState(true)
@@ -41,6 +44,16 @@ export function HistoryList({
 		setSessions((prev) => prev.filter((s) => s.id !== id))
 	}
 
+	const handleExport = (e: React.MouseEvent, session: SessionRecord) => {
+		e.stopPropagation()
+		downloadHistoryExport(session.task, session.createdAt, session.history)
+	}
+
+	const handleRerun = (e: React.MouseEvent, task: string) => {
+		e.stopPropagation()
+		onRerun(task)
+	}
+
 	return (
 		<div className="flex flex-col h-screen bg-background">
 			{/* Header */}
@@ -85,7 +98,6 @@ export function HistoryList({
 						role="button"
 						tabIndex={0}
 						onClick={() => onSelect(session.id)}
-						onKeyDown={(e) => e.key === 'Enter' && onSelect(session.id)}
 						className="w-full text-left px-3 py-2.5 border-b hover:bg-muted/50 transition-colors cursor-pointer flex items-start gap-2 group"
 					>
 						{/* Status icon */}
@@ -98,19 +110,41 @@ export function HistoryList({
 						{/* Content */}
 						<div className="flex-1 min-w-0">
 							<p className="text-xs font-medium truncate">{session.task}</p>
-							<p className="text-[10px] text-muted-foreground mt-0.5">
-								{timeAgo(session.createdAt)} · {session.history.length} steps
-							</p>
+							<div className="flex items-center mt-0.5">
+								<p className="text-[10px] text-muted-foreground">
+									{timeAgo(session.createdAt)} · {session.history.length} steps
+								</p>
+								<div className="flex items-center gap-0.5 ml-auto opacity-0 group-hover:opacity-100 transition-opacity">
+									<button
+										type="button"
+										onClick={(e) => handleRerun(e, session.task)}
+										className="p-0.5 text-muted-foreground hover:text-foreground transition-colors cursor-pointer"
+										title="Run task again"
+										aria-label={`Run history task again: ${session.task}`}
+									>
+										<RotateCcw className="size-3" />
+									</button>
+									<button
+										type="button"
+										onClick={(e) => handleExport(e, session)}
+										className="p-1 text-muted-foreground hover:text-foreground transition-colors cursor-pointer"
+										title="Export history JSON"
+										aria-label={`Export history for ${session.task}`}
+									>
+										<ArrowDownToLine className="size-3" />
+									</button>
+									<button
+										type="button"
+										onClick={(e) => handleDelete(e, session.id)}
+										className="p-0.5 text-muted-foreground hover:text-destructive transition-colors cursor-pointer"
+										title="Delete history"
+										aria-label={`Delete history for ${session.task}`}
+									>
+										<Trash2 className="size-3" />
+									</button>
+								</div>
+							</div>
 						</div>
-
-						{/* Delete */}
-						<button
-							type="button"
-							onClick={(e) => handleDelete(e, session.id)}
-							className="opacity-0 group-hover:opacity-100 transition-opacity p-1 hover:text-destructive cursor-pointer shrink-0"
-						>
-							<Trash2 className="size-3" />
-						</button>
 					</div>
 				))}
 			</div>

packages/extension/src/components/misc.tsx

@@ -84,7 +84,7 @@ export function MotionOverlay({ active }: { active: boolean }) {
 	return (
 		<div
 			ref={containerRef}
-			className="pointer-events-none absolute inset-0 z-10 opacity-60"
+			className="pointer-events-none absolute inset-0 z-10 opacity-60 overflow-hidden"
 			style={{ display: active ? undefined : 'none' }}
 		/>
 	)
@@ -111,6 +111,7 @@ export function EmptyState() {
 					]}
 					cursorStyle="underscore"
 					loop
+					startOnView={false}
 					typeSpeed={20}
 					deleteSpeed={10}
 					pauseDelay={3000}

packages/extension/src/components/ui/switch.tsx

@@ -8,7 +8,7 @@ function Switch({ className, ...props }: React.ComponentProps<typeof SwitchPrimi
 		<SwitchPrimitive.Root
 			data-slot="switch"
 			className={cn(
-				'peer data-[state=checked]:bg-primary data-[state=unchecked]:bg-input focus-visible:border-ring focus-visible:ring-ring/50 dark:data-[state=unchecked]:bg-input/80 inline-flex h-[1.15rem] w-8 shrink-0 items-center rounded-full border border-transparent shadow-xs transition-all outline-none focus-visible:ring-[3px] disabled:cursor-not-allowed disabled:opacity-50',
+				'peer data-[state=checked]:bg-primary data-[state=unchecked]:bg-input focus-visible:border-ring focus-visible:ring-ring/50 dark:data-[state=unchecked]:bg-input/80 inline-flex h-[1.15rem] w-8 shrink-0 cursor-pointer items-center rounded-full border border-transparent shadow-xs transition-all outline-none focus-visible:ring-[3px] disabled:cursor-not-allowed disabled:opacity-50',
 				className
 			)}
 			{...props}

packages/extension/src/entrypoints/background.ts

@@ -1,12 +1,12 @@
 import { handlePageControlMessage } from '@/agent/RemotePageController.background'
-import { handleTabControlMessage, setupTabChangeEvents } from '@/agent/TabsController.background'
+import { handleTabControlMessage, setupTabEventsPort } from '@/agent/TabsController.background'
 
 export default defineBackground(() => {
 	console.log('[Background] Service Worker started')
 
 	// tab change events
 
-	setupTabChangeEvents()
+	setupTabEventsPort()
 
 	// generate user auth token
 

packages/extension/src/entrypoints/content.ts

@@ -46,6 +46,8 @@ async function exposeAgentToPage() {
 	let multiPageAgent: InstanceType<typeof MultiPageAgent> | null = null
 
 	window.addEventListener('message', async (e) => {
+		if (e.source !== window) return
+
 		const data = e.data
 		if (typeof data !== 'object' || data === null) return
 		if (data.channel !== 'PAGE_AGENT_EXT_REQUEST') return
@@ -70,11 +72,15 @@ async function exposeAgentToPage() {
 
 				try {
 					const { task, config } = payload
+					const { systemInstruction, ...agentConfig } = config
 
 					// Dispose old instance before creating new one
 					multiPageAgent?.dispose()
 
-					multiPageAgent = new MultiPageAgent(config)
+					multiPageAgent = new MultiPageAgent({
+						...agentConfig,
+						instructions: systemInstruction ? { system: systemInstruction } : undefined,
+					})
 
 					// events
 

packages/extension/src/entrypoints/hub/App.tsx

@@ -1,10 +1,11 @@
-import { Plug, PlugZap, Square, Unplug } from 'lucide-react'
-import { useEffect, useRef } from 'react'
+import { FoldVertical, Plug, PlugZap, Square, UnfoldVertical, Unplug } from 'lucide-react'
+import { useEffect, useRef, useState } from 'react'
 
 import { useAgent } from '@/agent/useAgent'
 import { ActivityCard, EventCard } from '@/components/cards'
 import { Logo, MotionOverlay, StatusDot } from '@/components/misc'
 import { Button } from '@/components/ui/button'
+import { Switch } from '@/components/ui/switch'
 
 import { useHubWs } from './hub-ws'
 
@@ -25,33 +26,72 @@ export default function App() {
 	const wsLabel = {
 		connected: 'Connected',
 		connecting: 'Connecting…',
-		disconnected: new URLSearchParams(location.search).get('ws') ? 'Disconnected' : 'Standalone',
+		disconnected: new URLSearchParams(location.search).get('ws') ? 'Disconnected' : 'No connection',
 	}[wsState]
 
 	return (
 		<div className="flex h-screen bg-background">
-			<MotionOverlay active={isRunning} />
-
 			{/* Left — Protocol docs */}
 			<aside className="w-80 shrink-0 border-r flex flex-col bg-muted/20">
-				<div className="flex items-center gap-2 px-5 py-4 border-b">
+				<a
+					href="https://alibaba.github.io/page-agent/"
+					target="_blank"
+					rel="noopener noreferrer"
+					className="flex items-center gap-2 px-5 h-12 border-b hover:bg-muted/30 transition-colors"
+				>
 					<Logo className="size-5" />
 					<span className="text-sm font-semibold tracking-tight">Page Agent Hub</span>
+					<span className="text-[9px] font-medium uppercase tracking-wider text-amber-600 bg-amber-500/10 border border-amber-500/30 rounded px-1.5 py-0.5">
+						Beta
+					</span>
+				</a>
+
+				<div className="flex-1 overflow-y-auto px-5 py-4 space-y-6">
+					<div className="text-xs text-muted-foreground leading-relaxed space-y-2">
+						<p>
+							Page Agent Hub lets local apps (e.g. MCP servers) control the Page Agent extension via
+							WebSocket.
+						</p>
+						<p>
+							Check out the official{' '}
+							<a
+								href="https://github.com/alibaba/page-agent/tree/main/packages/mcp"
+								target="_blank"
+								rel="noopener noreferrer"
+								className="underline hover:text-foreground"
+							>
+								MCP server package
+							</a>
+							.
+						</p>
+					</div>
+
+					<HubConfig />
+
+					<ProtocolDocsCollapsible />
 				</div>
 
-				<div className="flex-1 overflow-y-auto px-5 py-4">
-					<ProtocolDocs />
-				</div>
-
-				<div className="border-t px-5 py-3 text-[11px] text-muted-foreground/60">
-					Connect via <code className="text-[10px]">hub.html?ws=PORT</code>
+				<div className="border-t px-5 py-3 text-[10px] text-muted-foreground/60 flex items-center justify-between">
+					<span className="font-mono">v{__VERSION__}</span>
+					<span>
+						Built with ♥️ by{' '}
+						<a
+							href="https://github.com/gaomeng1900"
+							target="_blank"
+							rel="noopener noreferrer"
+							className="underline hover:text-foreground"
+						>
+							@Simon
+						</a>
+					</span>
 				</div>
 			</aside>
 
 			{/* Right — Live session */}
-			<main className="flex-1 flex flex-col min-w-0">
-				{/* Header bar */}
-				<header className="flex items-center justify-between border-b px-5 py-3">
+			<main className="flex-1 flex flex-col min-w-0 relative">
+				<MotionOverlay active={isRunning} />
+
+				<header className="flex items-center justify-between border-b px-5 h-12">
 					<div className="flex items-center gap-2 text-xs text-muted-foreground">
 						<WsIcon className="size-3.5" />
 						<span>{wsLabel}</span>
@@ -104,49 +144,108 @@ export default function App() {
 	)
 }
 
-function ProtocolDocs() {
+function HubConfig() {
+	const [allowAll, setAllowAll] = useState(false)
+
+	useEffect(() => {
+		chrome.storage.local.get('allowAllHubConnection').then((r) => {
+			setAllowAll(r.allowAllHubConnection === true)
+		})
+	}, [])
+
+	const toggle = (checked: boolean) => {
+		setAllowAll(checked)
+		chrome.storage.local.set({ allowAllHubConnection: checked })
+	}
+
 	return (
-		<div className="space-y-5 text-xs text-muted-foreground">
-			<section>
-				<h3 className="text-[11px] font-semibold text-foreground/80 uppercase tracking-wider mb-2">
-					Caller → Hub
-				</h3>
-				<pre className="bg-muted/50 rounded-md p-3 font-mono text-[10px] leading-relaxed whitespace-pre-wrap">
-					{`{ type: "execute", task: string, config?: object }
-{ type: "stop" }`}
-				</pre>
-			</section>
+		<div>
+			<h3 className="text-[11px] font-semibold text-foreground/80 uppercase tracking-wider mb-2">
+				Config
+			</h3>
+			<div className="group/hub relative">
+				<label
+					className={`flex items-center justify-between p-3 rounded-md border cursor-pointer text-xs ${allowAll ? 'bg-amber-500/10 border-amber-500/30 text-amber-600' : 'bg-muted/50 text-muted-foreground'}`}
+				>
+					Auto-approve connections
+					<Switch
+						checked={allowAll}
+						onCheckedChange={toggle}
+						className={allowAll ? 'data-[state=checked]:bg-amber-500' : ''}
+					/>
+				</label>
 
-			<section>
-				<h3 className="text-[11px] font-semibold text-foreground/80 uppercase tracking-wider mb-2">
-					Hub → Caller
-				</h3>
-				<pre className="bg-muted/50 rounded-md p-3 font-mono text-[10px] leading-relaxed whitespace-pre-wrap">
-					{`{ type: "ready" }
-{ type: "result", success: boolean, data: string }
-{ type: "error", message: string }`}
-				</pre>
-			</section>
-
-			<section>
-				<h3 className="text-[11px] font-semibold text-foreground/80 uppercase tracking-wider mb-2">
-					Flow
-				</h3>
-				<ol className="list-decimal list-inside space-y-1 text-[11px] leading-relaxed">
-					<li>Hub opens WS to caller's server</li>
-					<li>
-						Sends <code className="text-[10px]">ready</code>
-					</li>
-					<li>
-						Caller sends <code className="text-[10px]">execute</code> with task
-					</li>
-					<li>Hub runs agent, streams events</li>
-					<li>
-						Hub sends <code className="text-[10px]">result</code> or{' '}
-						<code className="text-[10px]">error</code>
-					</li>
-				</ol>
-			</section>
+				{/* hide with invisible absolute opacity-0*/}
+				<div className="group-hover/hub:visible group-hover/hub:opacity-100 transition-opacity duration-150  left-0 right-0 top-full z-10 pt-2">
+					<div className="relative p-2.5 rounded-md border border-border bg-background/60 backdrop-blur-md shadow-2xl text-muted-foreground text-xs leading-relaxed">
+						<div className="absolute -top-1.5 left-5 size-3 rotate-45 rounded-[1px] border-l border-t border-border bg-background/60 backdrop-blur-md" />
+						By default, each connection requires your approval before running tasks. <br />
+						Enable this to skip per-session approval.
+						<br />
+						<span className="font-semibold">* Use with caution!</span>
+					</div>
+				</div>
+			</div>
+		</div>
+	)
+}
+
+function ProtocolDocsCollapsible() {
+	const [open, setOpen] = useState(false)
+
+	return (
+		<div>
+			<button
+				type="button"
+				onClick={() => setOpen(!open)}
+				className="flex items-center gap-1 text-[11px] font-semibold text-foreground/80 uppercase tracking-wider cursor-pointer"
+			>
+				Docs
+				{open ? <FoldVertical className="size-3" /> : <UnfoldVertical className="size-3" />}
+			</button>
+
+			{open && (
+				<div className="mt-3 space-y-4 text-xs text-muted-foreground">
+					<p className="text-[10px]">
+						Connect via <code className="text-[10px]">hub.html?ws=PORT</code>
+					</p>
+
+					<section>
+						<h4 className="text-[11px] font-medium text-foreground/60 mb-1.5">Flow</h4>
+						<ol className="list-decimal list-inside space-y-1 text-[11px] leading-relaxed">
+							<li>Hub opens WS to caller's server</li>
+							<li>
+								Sends <code className="text-[10px]">ready</code>
+							</li>
+							<li>
+								Caller sends <code className="text-[10px]">execute</code> with task
+							</li>
+							<li>Hub runs agent, streams events</li>
+							<li>
+								Hub sends <code className="text-[10px]">result</code> or{' '}
+								<code className="text-[10px]">error</code>
+							</li>
+						</ol>
+					</section>
+
+					<section>
+						<h4 className="text-[11px] font-medium text-foreground/60 mb-1.5">Caller → Hub</h4>
+						<pre className="bg-muted/50 rounded-md p-3 font-mono text-[10px] leading-relaxed whitespace-pre-wrap">
+							{`{ type: "execute", task: string, config?: object }
+{ type: "stop" }`}
+						</pre>
+					</section>
+
+					<section>
+						<h4 className="text-[11px] font-medium text-foreground/60 mb-1.5">Hub → Caller</h4>
+						<pre className="bg-muted/50 rounded-md p-3 font-mono text-[10px] leading-relaxed whitespace-pre-wrap">
+							{`{ type: "ready" }
+{ type: "result", success: boolean, data: string }
+{ type: "error", message: string }`}
+						</pre>
+					</section>
+				</div>
+			)}
 		</div>
 	)
 }

packages/extension/src/entrypoints/hub/hub-ws.ts

@@ -70,6 +70,7 @@ export class HubWs {
 	#ws: WebSocket | null = null
 	#state: HubWsState = 'disconnected'
 	#busy = false
+	#approved = false
 	#handlers: HubWsHandlers
 	#port: number
 	#onStateChange: (state: HubWsState) => void
@@ -103,6 +104,7 @@ export class HubWs {
 		ws.addEventListener('close', () => {
 			this.#ws = null
 			this.#busy = false
+			this.#approved = false
 			this.#setState('disconnected')
 		})
 
@@ -115,6 +117,7 @@ export class HubWs {
 		this.#ws?.close()
 		this.#ws = null
 		this.#busy = false
+		this.#approved = false
 		this.#setState('disconnected')
 	}
 
@@ -130,7 +133,7 @@ export class HubWs {
 		}
 	}
 
-	#handleMessage(raw: string) {
+	async #handleMessage(raw: string) {
 		let msg: InboundMessage
 		try {
 			msg = JSON.parse(raw)
@@ -138,6 +141,11 @@ export class HubWs {
 			return
 		}
 
+		if (!(await this.#checkApproval())) {
+			this.#send({ type: 'error', message: 'User denied the connection request.' })
+			return
+		}
+
 		switch (msg.type) {
 			case 'execute':
 				this.#handleExecute(msg)
@@ -148,6 +156,22 @@ export class HubWs {
 		}
 	}
 
+	async #checkApproval(): Promise<boolean> {
+		if (this.#approved) return true
+
+		const { allowAllHubConnection } = await chrome.storage.local.get('allowAllHubConnection')
+		if (allowAllHubConnection === true) {
+			this.#approved = true
+			return true
+		}
+
+		const ok = window.confirm(
+			'An external application is requesting to control your browser via Page Agent Ext.\nAllow this session?'
+		)
+		if (ok) this.#approved = true
+		return ok
+	}
+
 	async #handleExecute(msg: ExecuteMessage) {
 		if (this.#busy) {
 			this.#send({ type: 'error', message: 'Hub is busy with another task' })

packages/extension/src/entrypoints/main-world.ts

@@ -4,8 +4,14 @@ export type Execute = (task: string, config: ExecuteConfig) => Promise<Execution
 
 export interface ExecuteConfig {
 	baseURL: string
-	apiKey: string
 	model: string
+	apiKey?: string
+
+	/**
+	 * Global system-level instructions for the agent.
+	 * Equivalent to `AgentConfig.instructions.system`.
+	 */
+	systemInstruction?: string
 
 	/**
 	 * Whether to include the initial tab (that holds this main world script) in the task.
@@ -13,6 +19,9 @@ export interface ExecuteConfig {
 	 */
 	includeInitialTab?: boolean
 
+	/** Control all unpinned tabs in the window instead of only the tab group. */
+	experimentalIncludeAllTabs?: boolean
+
 	onStatusChange?: (status: AgentStatus) => void
 	onActivity?: (activity: AgentActivity) => void
 	onHistoryUpdate?: (history: HistoricalEvent[]) => void
@@ -30,13 +39,14 @@ export default defineUnlistedScript(() => {
 		if (task.trim().length === 0) throw new Error('Task cannot be empty')
 		if (!config) throw new Error('Config is required')
 		if (!config.baseURL) throw new Error('Config must have a baseURL')
-		if (!config.apiKey) throw new Error('Config must have an apiKey')
 		if (!config.model) throw new Error('Config must have a model')
 
 		const id = getId()
 
 		const promise = new Promise<ExecutionResult>((resolve, reject) => {
 			function handleMessage(e: MessageEvent) {
+				if (e.source !== window) return
+
 				const data = e.data
 				if (typeof data !== 'object' || data === null) return
 				if (data.channel !== 'PAGE_AGENT_EXT_RESPONSE') return
@@ -85,9 +95,11 @@ export default defineUnlistedScript(() => {
 					task,
 					config: {
 						baseURL: config.baseURL,
-						apiKey: config.apiKey,
 						model: config.model,
+						apiKey: config.apiKey,
+						systemInstruction: config.systemInstruction,
 						includeInitialTab: config.includeInitialTab,
+						experimentalIncludeAllTabs: config.experimentalIncludeAllTabs,
 					},
 				},
 			},

packages/extension/src/entrypoints/sidepanel/App.tsx

@@ -56,19 +56,27 @@ export default function App() {
 		}
 	}, [history, activity])
 
-	const handleSubmit = useCallback(
-		(e?: React.SyntheticEvent) => {
-			e?.preventDefault()
-			if (!inputValue.trim() || status === 'running') return
+	const runTask = useCallback(
+		(task: string) => {
+			const normalizedTask = task.trim()
+			if (!normalizedTask || status === 'running') return
 
-			const taskToExecute = inputValue.trim()
 			setInputValue('')
+			setView({ name: 'chat' })
 
-			execute(taskToExecute).catch((error) => {
+			execute(normalizedTask).catch((error) => {
 				console.error('[SidePanel] Failed to execute task:', error)
 			})
 		},
-		[inputValue, status, execute]
+		[execute, status]
+	)
+
+	const handleSubmit = useCallback(
+		(e?: React.SyntheticEvent) => {
+			e?.preventDefault()
+			runTask(inputValue)
+		},
+		[inputValue, runTask]
 	)
 
 	const handleStop = useCallback(() => {
@@ -103,12 +111,19 @@ export default function App() {
 			<HistoryList
 				onSelect={(id) => setView({ name: 'history-detail', sessionId: id })}
 				onBack={() => setView({ name: 'chat' })}
+				onRerun={runTask}
 			/>
 		)
 	}
 
 	if (view.name === 'history-detail') {
-		return <HistoryDetail sessionId={view.sessionId} onBack={() => setView({ name: 'history' })} />
+		return (
+			<HistoryDetail
+				sessionId={view.sessionId}
+				onBack={() => setView({ name: 'history' })}
+				onRerun={runTask}
+			/>
+		)
 	}
 
 	// --- Chat view ---

packages/extension/src/lib/history-export.ts

@@ -0,0 +1,60 @@
+import type { HistoricalEvent } from '@page-agent/core'
+
+const EXPORT_FILE_PREFIX = 'page-agent-history'
+const MAX_TASK_SLUG_LENGTH = 40
+
+export function serializeHistoryExport(history: HistoricalEvent[]): string {
+	return `${JSON.stringify(history, null, 2)}\n`
+}
+
+export function buildHistoryExportFilename(task: string, createdAt: number): string {
+	const taskSlug = sanitizeTaskForFilename(task)
+	const timestamp = formatTimestampForFilename(createdAt)
+
+	return taskSlug
+		? `${EXPORT_FILE_PREFIX}-${taskSlug}-${timestamp}.json`
+		: `${EXPORT_FILE_PREFIX}-${timestamp}.json`
+}
+
+export function downloadHistoryExport(
+	task: string,
+	createdAt: number,
+	history: HistoricalEvent[]
+): void {
+	const filename = buildHistoryExportFilename(task, createdAt)
+	const content = serializeHistoryExport(history)
+	const blob = new Blob([content], { type: 'application/json;charset=utf-8' })
+	const url = URL.createObjectURL(blob)
+	const link = document.createElement('a')
+
+	link.href = url
+	link.download = filename
+	link.click()
+
+	URL.revokeObjectURL(url)
+}
+
+function sanitizeTaskForFilename(task: string): string {
+	return task
+		.trim()
+		.toLowerCase()
+		.replace(/[^a-z0-9]+/g, '-')
+		.replace(/^-+|-+$/g, '')
+		.slice(0, MAX_TASK_SLUG_LENGTH)
+}
+
+function formatTimestampForFilename(createdAt: number): string {
+	const date = new Date(createdAt)
+	const year = date.getFullYear()
+	const month = pad(date.getMonth() + 1)
+	const day = pad(date.getDate())
+	const hours = pad(date.getHours())
+	const minutes = pad(date.getMinutes())
+	const seconds = pad(date.getSeconds())
+
+	return `${year}-${month}-${day}_${hours}-${minutes}-${seconds}`
+}
+
+function pad(value: number): string {
+	return value.toString().padStart(2, '0')
+}

packages/llms/package.json

@@ -1,6 +1,6 @@
 {
     "name": "@page-agent/llms",
-    "version": "1.5.9",
+    "version": "1.7.1",
     "type": "module",
     "main": "./dist/lib/page-agent-llms.js",
     "module": "./dist/lib/page-agent-llms.js",

packages/llms/src/OpenAIClient.ts

@@ -29,16 +29,19 @@ export class OpenAIClient implements LLMClient {
 		const openaiTools = Object.entries(tools).map(([name, t]) => zodToOpenAITool(name, t))
 
 		// Build request body
+
+		let toolChoice: unknown = 'required'
+		if (options?.toolChoiceName && !this.config.disableNamedToolChoice) {
+			toolChoice = { type: 'function', function: { name: options.toolChoiceName } }
+		}
+
 		const requestBody: Record<string, unknown> = {
 			model: this.config.model,
 			temperature: this.config.temperature,
 			messages,
 			tools: openaiTools,
 			parallel_tool_calls: false,
-			// Require tool call: specific tool if provided, otherwise any tool
-			tool_choice: options?.toolChoiceName
-				? { type: 'function', function: { name: options.toolChoiceName } }
-				: 'required',
+			tool_choice: toolChoice,
 		}
 
 		modelPatch(requestBody)
@@ -50,7 +53,7 @@ export class OpenAIClient implements LLMClient {
 				method: 'POST',
 				headers: {
 					'Content-Type': 'application/json',
-					Authorization: `Bearer ${this.config.apiKey}`,
+					...(this.config.apiKey && { Authorization: `Bearer ${this.config.apiKey}` }),
 				},
 				body: JSON.stringify(requestBody),
 				signal: abortSignal,

packages/llms/src/index.ts

@@ -8,19 +8,20 @@ export type { InvokeOptions, InvokeResult, LLMClient, LLMConfig, Message, Tool }
 
 export function parseLLMConfig(config: LLMConfig): Required<LLMConfig> {
 	// Runtime validation as defensive programming (types already guarantee these)
-	if (!config.baseURL || !config.apiKey || !config.model) {
+	if (!config.baseURL || !config.model) {
 		throw new Error(
-			'[PageAgent] LLM configuration required. Please provide: baseURL, apiKey, model. ' +
+			'[PageAgent] LLM configuration required. Please provide: baseURL, model. ' +
 				'See: https://alibaba.github.io/page-agent/docs/features/models'
 		)
 	}
 
 	return {
 		baseURL: config.baseURL,
-		apiKey: config.apiKey,
 		model: config.model,
+		apiKey: config.apiKey || '',
 		temperature: config.temperature ?? DEFAULT_TEMPERATURE,
 		maxRetries: config.maxRetries ?? LLM_MAX_RETRIES,
+		disableNamedToolChoice: config.disableNamedToolChoice ?? false,
 		customFetch: (config.customFetch ?? fetch).bind(globalThis), // fetch will be illegal unless bound
 	}
 }

packages/llms/src/types.ts

@@ -89,12 +89,18 @@ export interface InvokeResult<TResult = unknown> {
  */
 export interface LLMConfig {
 	baseURL: string
-	apiKey: string
 	model: string
+	apiKey?: string
 
 	temperature?: number
 	maxRetries?: number
 
+	/**
+	 * remove the tool_choice field from the request.
+	 * @note fix "Invalid tool_choice type: 'object'" for some LLMs.
+	 */
+	disableNamedToolChoice?: boolean
+
 	/**
 	 * Custom fetch function for LLM API requests.
 	 * Use this to customize headers, credentials, proxy, etc.

packages/mcp/package.json

@@ -1,7 +1,7 @@
 {
     "name": "@page-agent/mcp",
-    "private": true,
-    "version": "1.5.9",
+    "private": false,
+    "version": "1.7.1",
     "type": "module",
     "bin": {
         "page-agent-mcp": "src/index.js"
@@ -28,8 +28,8 @@
         "node": ">=20"
     },
     "dependencies": {
-        "@modelcontextprotocol/sdk": "^1.27.1",
-        "ws": "^8.19.0",
+        "@modelcontextprotocol/sdk": "^1.29.0",
+        "ws": "^8.20.0",
         "zod": "^4.3.5"
     }
 }

packages/mcp/src/index.js

@@ -12,9 +12,9 @@ const port = parseInt(env.PORT || '38401')
 
 /** @type {Record<string, string>} */
 const llmConfig = {}
-if (env.LLM_API_KEY) llmConfig.apiKey = env.LLM_API_KEY
 if (env.LLM_BASE_URL) llmConfig.baseURL = env.LLM_BASE_URL
 if (env.LLM_MODEL_NAME) llmConfig.model = env.LLM_MODEL_NAME
+if (env.LLM_API_KEY) llmConfig.apiKey = env.LLM_API_KEY
 
 // --- Hub bridge (HTTP + WebSocket) ---
 
@@ -35,11 +35,14 @@ const mcpServer = new McpServer({ name: 'page-agent', version: '1.5.8' })
 mcpServer.registerTool(
 	'execute_task',
 	{
-		description:
-			'Execute a browser automation task described in natural language. ' +
-			'The Page Agent extension will control the browser to complete the task. ' +
-			'Blocks until the task is complete.',
-		inputSchema: { task: z.string().describe('Task description in natural language') },
+		description: "Execute a task in user's browser.",
+		inputSchema: {
+			task: z
+				.string()
+				.describe(
+					'Task description. Give specific instructions for the task. Steps preferable. And the information you want to get after the task is done.'
+				),
+		},
 	},
 	async ({ task }) => {
 		try {
@@ -50,7 +53,7 @@ mcpServer.registerTool(
 					{
 						type: 'text',
 						text: result.success
-							? `Task completed successfully.\n\n${result.data}`
+							? `Task completed.\n\n${result.data}`
 							: `Task failed.\n\n${result.data}`,
 					},
 				],
@@ -67,7 +70,7 @@ mcpServer.registerTool(
 mcpServer.registerTool(
 	'get_status',
 	{
-		description: 'Check the current status of the Page Agent hub connection and agent.',
+		description: 'Check the current status of the Page Agent hub.',
 	},
 	async () => ({
 		content: [

packages/mcp/src/launcher.html

@@ -1,9 +1,10 @@
 <!doctype html>
-<html lang="en">
+<html>
 	<head>
 		<meta charset="UTF-8" />
 		<meta name="viewport" content="width=device-width, initial-scale=1.0" />
-		<title>Page Agent – Connecting</title>
+		<link rel="icon" href="https://img.alicdn.com/imgextra/i1/O1CN01mRGret1QrKiu7CFJI_!!6000000002029-2-tps-64-64.png" />
+		<title>Page Agent MCP Launcher</title>
 		<style>
 			* {
 				margin: 0;
@@ -15,7 +16,7 @@
 					system-ui,
 					-apple-system,
 					sans-serif;
-				background: #0a0a0a;
+				background: #09090b;
 				color: #e5e5e5;
 				display: flex;
 				align-items: center;
@@ -24,23 +25,37 @@
 			}
 			.card {
 				text-align: center;
-				max-width: 420px;
+				max-width: 480px;
 				padding: 3rem 2rem;
 			}
+			.logo {
+				width: 72px;
+				height: 72px;
+				border-radius: 18px;
+				margin-bottom: 1rem;
+			}
+			.badge {
+				font-size: 0.6875rem;
+				font-weight: 500;
+				color: #52525b;
+				letter-spacing: 0.05em;
+				text-transform: uppercase;
+				margin-bottom: 1.5rem;
+			}
 			h1 {
-				font-size: 1.25rem;
+				font-size: 1.35rem;
 				font-weight: 600;
 				margin-bottom: 0.5rem;
 			}
-			p {
+			.sub {
 				font-size: 0.875rem;
-				color: #a3a3a3;
-				line-height: 1.6;
+				color: #a1a1aa;
+				line-height: 1.7;
 			}
 			.spinner {
 				width: 32px;
 				height: 32px;
-				border: 3px solid #333;
+				border: 3px solid #27272a;
 				border-top-color: #fff;
 				border-radius: 50%;
 				animation: spin 0.8s linear infinite;
@@ -58,72 +73,165 @@
 			a:hover {
 				text-decoration: underline;
 			}
+
 			.install {
 				display: none;
 			}
 			.install.show {
 				display: block;
 			}
-			.btn {
-				display: inline-block;
-				margin-top: 1rem;
-				padding: 0.5rem 1.25rem;
+
+			.tips {
+				margin-top: 1.75rem;
+				text-align: left;
+				background: #18181b;
+				border: 1px solid #27272a;
+				border-radius: 12px;
+				padding: 1.25rem 1.5rem;
+			}
+			.tips li {
+				font-size: 0.8125rem;
+				color: #a1a1aa;
+				line-height: 1.7;
+				margin-left: 1rem;
+			}
+			.tips li + li {
+				margin-top: 0.35rem;
+			}
+
+			.store-btn {
+				display: inline-flex;
+				align-items: center;
+				gap: 0.625rem;
+				margin-top: 1.5rem;
+				padding: 0.625rem 1.5rem;
 				background: #2563eb;
 				color: #fff;
-				border-radius: 0.5rem;
+				border-radius: 10px;
 				font-size: 0.875rem;
 				font-weight: 500;
+				transition: background 0.15s;
 			}
-			.btn:hover {
+			.store-btn:hover {
 				background: #1d4ed8;
 				text-decoration: none;
 			}
+			.store-btn img {
+				width: 20px;
+				height: 20px;
+			}
+
+			.links {
+				margin-top: 1.75rem;
+				display: flex;
+				justify-content: center;
+				gap: 1.5rem;
+				font-size: 0.8125rem;
+			}
+			.links a {
+				color: #71717a;
+				transition: color 0.15s;
+			}
+			.links a:hover {
+				color: #a1a1aa;
+				text-decoration: none;
+			}
 		</style>
 	</head>
 	<body>
 		<div class="card">
+			<img
+				class="logo"
+				src="https://img.alicdn.com/imgextra/i3/O1CN01JPT4Fj1FJTfmHfNxO_!!6000000000466-49-tps-512-512.webp"
+				alt="Page Agent"
+			/>
+			<div class="badge">Page Agent MCP Launcher</div>
+
 			<div id="connecting">
 				<div class="spinner"></div>
-				<h1>Connecting to Page Agent</h1>
-				<p>Opening the hub in your browser...</p>
+				<h1 data-i18n="connecting_title">Connecting to Page Agent</h1>
+				<p class="sub" data-i18n="connecting_sub">Opening the hub in your browser…</p>
 			</div>
+
 			<div id="install" class="install">
-				<h1>Extension Not Found</h1>
-				<p>The Page Agent browser extension is required.<br />Install it and try again.</p>
-				<a class="btn" href="__STORE_URL__" target="_blank">Install Extension</a>
-				<p style="margin-top: 1.5rem; font-size: 0.75rem">
-					After installing, restart the MCP server or refresh this page.
+				<h1 data-i18n="install_title">Extension Required</h1>
+				<p class="sub" data-i18n="install_sub">
+					Page Agent requires the latest browser extension to work.
 				</p>
+
+				<a class="store-btn" href="__STORE_URL__" target="_blank">
+					<img
+						src="https://img.alicdn.com/imgextra/i3/O1CN01JpW0Vo1sR3FpiZKFM_!!6000000005762-55-tps-192-192.svg"
+						alt=""
+					/>
+					<span data-i18n="install_btn">Install from Chrome Web Store</span>
+				</a>
+
+				<ul class="tips">
+					<li data-i18n="tip_outdated">
+						If the extension is outdated, please update it to the latest version.
+					</li>
+					<li data-i18n="tip_other_browser">
+						If the extension is not installed in this browser, open this page from the
+						browser that has it installed.
+					</li>
+					<li data-i18n="tip_refresh">Refresh this page after installing or updating.</li>
+				</ul>
+			</div>
+
+			<div class="links">
+				<a href="https://alibaba.github.io/page-agent/docs/introduction/overview" target="_blank" data-i18n="link_docs">Docs</a>
+				<a href="https://github.com/alibaba/page-agent/issues" target="_blank" data-i18n="link_issues">Report an Issue</a>
 			</div>
 		</div>
-		<script>
-			;(function () {
-				var EXT_ID = '__EXT_ID__'
-				var wsPort = __WS_PORT__
 
-				function showInstall() {
+		<script>
+			{
+				const EXT_ID = '__EXT_ID__'
+				const wsPort = __WS_PORT__
+
+				const zh = {
+					connecting_title: '正在连接 Page Agent',
+					connecting_sub: '正在浏览器中打开 Hub…',
+					install_title: '需要安装浏览器插件',
+					install_sub: 'Page Agent 需要安装最新版浏览器插件才能运行。',
+					install_btn: '从 Chrome 应用商店安装',
+					tip_outdated: '如果插件版本过旧，请更新到最新版本。',
+					tip_other_browser:
+						'如果该浏览器中未安装插件，请从装有插件的浏览器打开此页面。',
+					tip_refresh: '安装或更新后，请刷新此页面。',
+					link_docs: '文档',
+					link_issues: '问题反馈',
+				}
+
+				if (/^zh\b/i.test(navigator.language)) {
+					document.querySelectorAll('[data-i18n]').forEach((el) => {
+						const key = el.getAttribute('data-i18n')
+						if (zh[key]) el.textContent = zh[key]
+					})
+				}
+
+				const showInstall = () => {
 					document.getElementById('connecting').style.display = 'none'
 					document.getElementById('install').classList.add('show')
 				}
 
 				try {
-					if (typeof chrome === 'undefined' || !chrome.runtime || !chrome.runtime.sendMessage) {
+					if (!globalThis.chrome?.runtime?.sendMessage) {
 						showInstall()
-						return
-					}
-					chrome.runtime.sendMessage(
-						EXT_ID,
-						{ type: 'OPEN_HUB', wsPort: wsPort },
-						function (response) {
-							if (chrome.runtime.lastError || !response || !response.ok) {
-								showInstall()
+					} else {
+						chrome.runtime.sendMessage(
+							EXT_ID,
+							{ type: 'OPEN_HUB', wsPort },
+							(response) => {
+								if (chrome.runtime.lastError || !response?.ok) showInstall()
 							}
-						}
-					)
-				} catch (e) {
+						)
+					}
+				} catch {
 					showInstall()
 				}
-			})()
+			}
 		</script>
 	</body>
 </html>

packages/page-agent/package.json

@@ -1,7 +1,7 @@
 {
     "name": "page-agent",
     "private": false,
-    "version": "1.5.9",
+    "version": "1.7.1",
     "type": "module",
     "main": "./dist/esm/page-agent.js",
     "module": "./dist/esm/page-agent.js",
@@ -44,10 +44,10 @@
         "postpublish": "node -e \"['README.md','LICENSE'].forEach(f=>{try{require('fs').unlinkSync(f)}catch{}})\""
     },
     "dependencies": {
-        "@page-agent/core": "1.5.9",
-        "@page-agent/llms": "1.5.9",
-        "@page-agent/page-controller": "1.5.9",
-        "@page-agent/ui": "1.5.9",
+        "@page-agent/core": "1.7.1",
+        "@page-agent/llms": "1.7.1",
+        "@page-agent/page-controller": "1.7.1",
+        "@page-agent/ui": "1.7.1",
         "chalk": "^5.6.2"
     },
     "peerDependencies": {

packages/page-agent/src/PageAgent.ts

@@ -4,11 +4,11 @@
  */
 import { type AgentConfig, PageAgentCore } from '@page-agent/core'
 import { PageController, type PageControllerConfig } from '@page-agent/page-controller'
-import { Panel } from '@page-agent/ui'
+import { Panel, type PanelConfig } from '@page-agent/ui'
 
 export * from '@page-agent/core'
 
-export type PageAgentConfig = AgentConfig & PageControllerConfig
+export type PageAgentConfig = AgentConfig & PageControllerConfig & Omit<PanelConfig, 'language'>
 
 export class PageAgent extends PageAgentCore {
 	panel: Panel
@@ -23,6 +23,7 @@ export class PageAgent extends PageAgentCore {
 
 		this.panel = new Panel(this, {
 			language: config.language,
+			promptForNextTask: config.promptForNextTask,
 		})
 	}
 }

packages/page-agent/src/demo.ts

@@ -17,9 +17,10 @@ const DEMO_MODEL = 'qwen3.5-plus'
 const DEMO_BASE_URL = 'https://page-ag-testing-ohftxirgbn.cn-shanghai.fcapp.run'
 const DEMO_API_KEY = 'NA'
 
+const currentScript = document.currentScript as HTMLScriptElement | null
+
 // in case document.x is not ready yet
 setTimeout(() => {
-	const currentScript = document.currentScript as HTMLScriptElement | null
 	let config: PageAgentConfig
 
 	if (currentScript) {

packages/page-controller/package.json

@@ -1,6 +1,6 @@
 {
     "name": "@page-agent/page-controller",
-    "version": "1.5.9",
+    "version": "1.7.1",
     "type": "module",
     "main": "./dist/lib/page-controller.js",
     "module": "./dist/lib/page-controller.js",

packages/page-controller/src/PageController.ts

@@ -18,6 +18,7 @@ import * as dom from './dom'
 import type { FlatDomTree, InteractiveElementDomNode } from './dom/dom_tree/type'
 import { getPageInfo } from './dom/getPageInfo'
 import { patchReact } from './patches/react'
+import { isAnchorElement } from './utils'
 
 /**
  * Configuration for PageController
@@ -192,7 +193,11 @@ export class PageController extends EventTarget {
 			interactiveBlacklist: blacklist,
 		})
 
-		this.simplifiedHTML = dom.flatTreeToString(this.flatTree, this.config.includeAttributes)
+		this.simplifiedHTML = dom.flatTreeToString(
+			this.flatTree,
+			this.config.includeAttributes,
+			this.config.keepSemanticTags
+		)
 
 		this.selectorMap.clear()
 		this.selectorMap = dom.getSelectorMap(this.flatTree)
@@ -217,6 +222,7 @@ export class PageController extends EventTarget {
 	 * Clean up all element highlights
 	 */
 	async cleanUpHighlights(): Promise<void> {
+		console.log('[PageController] cleanUpHighlights')
 		dom.cleanUpHighlights()
 	}
 
@@ -243,7 +249,7 @@ export class PageController extends EventTarget {
 			await clickElement(element)
 
 			// Handle links that open in new tabs
-			if (element instanceof HTMLAnchorElement && element.target === '_blank') {
+			if (isAnchorElement(element) && element.target === '_blank') {
 				return {
 					success: true,
 					message: `✅ Clicked element (${elemText ?? index}). ⚠️ Link opened in a new tab.`,
@@ -320,11 +326,11 @@ export class PageController extends EventTarget {
 
 			this.assertIndexed()
 
-			const scrollAmount = pixels ?? numPages * (down ? 1 : -1) * window.innerHeight
+			const scrollAmount = (pixels ?? numPages * window.innerHeight) * (down ? 1 : -1)
 
 			const element = index !== undefined ? getElementByIndex(this.selectorMap, index) : null
 
-			const message = await scrollVertically(down, scrollAmount, element)
+			const message = await scrollVertically(scrollAmount, element)
 
 			return {
 				success: true,
@@ -355,7 +361,7 @@ export class PageController extends EventTarget {
 
 			const element = index !== undefined ? getElementByIndex(this.selectorMap, index) : null
 
-			const message = await scrollHorizontally(right, scrollAmount, element)
+			const message = await scrollHorizontally(scrollAmount, element)
 
 			return {
 				success: true,
@@ -423,3 +429,5 @@ export class PageController extends EventTarget {
 		this.mask = null
 	}
 }
+
+export * from './actions'

packages/page-controller/src/actions.ts

@@ -3,27 +3,22 @@
  * All rights reserved.
  */
 import type { InteractiveElementDomNode } from './dom/dom_tree/type'
-
-// ======= general utils =======
-
-async function waitFor(seconds: number): Promise<void> {
-	await new Promise((resolve) => setTimeout(resolve, seconds * 1000))
-}
-
-// ======= dom utils =======
-
-export async function movePointerToElement(element: HTMLElement) {
-	const rect = element.getBoundingClientRect()
-	const x = rect.left + rect.width / 2
-	const y = rect.top + rect.height / 2
-
-	window.dispatchEvent(new CustomEvent('PageAgent::MovePointerTo', { detail: { x, y } }))
-
-	await waitFor(0.3)
-}
+import {
+	clickPointer,
+	disablePassThrough,
+	enablePassThrough,
+	getNativeValueSetter,
+	isHTMLElement,
+	isInputElement,
+	isSelectElement,
+	isTextAreaElement,
+	movePointerToElement,
+	waitFor,
+} from './utils'
 
 /**
  * Get the HTMLElement by index from a selectorMap.
+ * @private Internal method, subject to change at any time.
  */
 export function getElementByIndex(
 	selectorMap: Map<number, InteractiveElementDomNode>,
@@ -39,7 +34,7 @@ export function getElementByIndex(
 		throw new Error(`Element at index ${index} does not have a reference`)
 	}
 
-	if (!(element instanceof HTMLElement)) {
+	if (!isHTMLElement(element)) {
 		throw new Error(`Element at index ${index} is not an HTMLElement`)
 	}
 
@@ -50,64 +45,92 @@ let lastClickedElement: HTMLElement | null = null
 
 function blurLastClickedElement() {
 	if (lastClickedElement) {
+		lastClickedElement.dispatchEvent(new PointerEvent('pointerout', { bubbles: true }))
+		lastClickedElement.dispatchEvent(new PointerEvent('pointerleave', { bubbles: false }))
+		lastClickedElement.dispatchEvent(new MouseEvent('mouseout', { bubbles: true }))
+		lastClickedElement.dispatchEvent(new MouseEvent('mouseleave', { bubbles: false }))
 		lastClickedElement.blur()
-		lastClickedElement.dispatchEvent(
-			new MouseEvent('mouseout', { bubbles: true, cancelable: true })
-		)
 		lastClickedElement = null
 	}
 }
 
 /**
- * Simulate a click on the element
+ * Simulate a full click following W3C Pointer Events + UI Events spec order:
+ * pointerover/enter → mouseover/enter → pointerdown → mousedown → [focus] →
+ * pointerup → mouseup → click
+ *
+ * @private Internal method, subject to change at any time.
  */
 export async function clickElement(element: HTMLElement) {
 	blurLastClickedElement()
 
 	lastClickedElement = element
+
 	await scrollIntoViewIfNeeded(element)
-	await movePointerToElement(element)
-	window.dispatchEvent(new CustomEvent('PageAgent::ClickPointer'))
+	const frame = element.ownerDocument.defaultView?.frameElement
+	if (frame) await scrollIntoViewIfNeeded(frame)
+
+	const rect = element.getBoundingClientRect()
+	const x = rect.left + rect.width / 2
+	const y = rect.top + rect.height / 2
+
+	await movePointerToElement(element, x, y)
+	await clickPointer()
+
 	await waitFor(0.1)
 
-	// hover it
-	element.dispatchEvent(new MouseEvent('mouseenter', { bubbles: true, cancelable: true }))
-	element.dispatchEvent(new MouseEvent('mouseover', { bubbles: true, cancelable: true }))
+	// Hit-test to find the deepest element at click coordinates, matching
+	// real browser behavior where events target the innermost element.
+	// @note This may hit a element in the blacklist
+	// TODO: This is a temporary workaround. Should have been handled during dom extraction.
+	const doc = element.ownerDocument
+	await enablePassThrough()
+	const hitTarget = doc.elementFromPoint(x, y)
+	await disablePassThrough()
+	const target =
+		hitTarget instanceof HTMLElement && element.contains(hitTarget) ? hitTarget : element
 
-	// dispatch a sequence of events to ensure all listeners are triggered
-	element.dispatchEvent(new MouseEvent('mousedown', { bubbles: true, cancelable: true }))
+	const pointerOpts = {
+		bubbles: true,
+		cancelable: true,
+		clientX: x,
+		clientY: y,
+		pointerType: 'mouse',
+	}
+	const mouseOpts = { bubbles: true, cancelable: true, clientX: x, clientY: y, button: 0 }
 
-	// focus it to ensure it gets the click event
-	element.focus()
+	// Hover — pointer events first, then mouse events (spec order)
+	target.dispatchEvent(new PointerEvent('pointerover', pointerOpts))
+	target.dispatchEvent(new PointerEvent('pointerenter', { ...pointerOpts, bubbles: false }))
+	target.dispatchEvent(new MouseEvent('mouseover', mouseOpts))
+	target.dispatchEvent(new MouseEvent('mouseenter', { ...mouseOpts, bubbles: false }))
 
-	element.dispatchEvent(new MouseEvent('mouseup', { bubbles: true, cancelable: true }))
-	element.dispatchEvent(new MouseEvent('click', { bubbles: true, cancelable: true }))
+	// Press
+	target.dispatchEvent(new PointerEvent('pointerdown', pointerOpts))
+	target.dispatchEvent(new MouseEvent('mousedown', mouseOpts))
 
-	// dispatch a click event
-	// element.click()
+	// Focus is not part of the standard pointer/mouse event sequence
+	// "undefined and varies between user agents".
+	// We focus the original element (nearest focusable ancestor), not the hit-test target, matching browser behavior.
+	element.focus({ preventScroll: true })
 
-	await waitFor(0.2) // Wait to ensure click event processing completes
+	// Release
+	target.dispatchEvent(new PointerEvent('pointerup', pointerOpts))
+	target.dispatchEvent(new MouseEvent('mouseup', mouseOpts))
+
+	// Click — activation behavior (navigation, form submit, etc.) triggers
+	// via bubbling from target up to the interactive ancestor.
+	target.click()
+
+	await waitFor(0.2)
 }
 
-// eslint-disable-next-line @typescript-eslint/unbound-method
-const nativeInputValueSetter = Object.getOwnPropertyDescriptor(
-	window.HTMLInputElement.prototype,
-	'value'
-)!.set!
-
-// eslint-disable-next-line @typescript-eslint/unbound-method
-const nativeTextAreaValueSetter = Object.getOwnPropertyDescriptor(
-	window.HTMLTextAreaElement.prototype,
-	'value'
-)!.set!
-
+/**
+ * @private Internal method, subject to change at any time.
+ */
 export async function inputTextElement(element: HTMLElement, text: string) {
 	const isContentEditable = element.isContentEditable
-	if (
-		!(element instanceof HTMLInputElement) &&
-		!(element instanceof HTMLTextAreaElement) &&
-		!isContentEditable
-	) {
+	if (!isInputElement(element) && !isTextAreaElement(element) && !isContentEditable) {
 		throw new Error('Element is not an input, textarea, or contenteditable')
 	}
 
@@ -119,9 +142,12 @@ export async function inputTextElement(element: HTMLElement, text: string) {
 		// - Monaco/CodeMirror: Require direct JS instance access. No universal way to obtain.
 		// - Draft.js: Not responsive to synthetic/execCommand/Range/DataTransfer. Unmaintained.
 		//
+		// Strategy: Try Plan A (synthetic events) first, then verify and fall back
+		// to Plan B (execCommand) if the text wasn't actually inserted.
+		//
 		// Plan A: Dispatch synthetic events
-		// Works: LinkedIn, React contenteditable, Quill.
-		// Fails: Slate.js
+		// Works: React contenteditable, Quill.
+		// Fails: Slate.js, some contenteditable editors that ignore synthetic events.
 		// Sequence: beforeinput -> mutation -> input -> change -> blur
 
 		// Dispatch beforeinput + mutation + input for clearing
@@ -164,22 +190,37 @@ export async function inputTextElement(element: HTMLElement, text: string) {
 			)
 		}
 
+		// Verify Plan A worked by checking if the text was actually inserted
+		const planASucceeded = element.innerText.trim() === text.trim()
+
+		if (!planASucceeded) {
+			// Plan B: execCommand fallback (deprecated but widely supported)
+			// Works: Quill, Slate.js, react contenteditable components.
+			// This approach integrates with the browser's undo stack and is handled
+			// natively by most rich-text editors.
+			element.focus()
+
+			// Select all existing content and delete it
+			const doc = element.ownerDocument
+			const selection = (doc.defaultView || window).getSelection()
+			const range = doc.createRange()
+			range.selectNodeContents(element)
+			selection?.removeAllRanges()
+			selection?.addRange(range)
+
+			// eslint-disable-next-line @typescript-eslint/no-deprecated
+			doc.execCommand('delete', false)
+			// eslint-disable-next-line @typescript-eslint/no-deprecated
+			doc.execCommand('insertText', false, text)
+		}
+
 		// Dispatch change event (for good measure)
 		element.dispatchEvent(new Event('change', { bubbles: true }))
 
 		// Trigger blur for validation
 		element.blur()
-
-		// Plan B: execCommand (deprecated but works better for some editors)
-		// Works: LinkedIn, Quill, Slate.js, react contenteditable components
-		//
-		// document.execCommand('selectAll')
-		// document.execCommand('delete')
-		// document.execCommand('insertText', false, text)
-	} else if (element instanceof HTMLTextAreaElement) {
-		nativeTextAreaValueSetter.call(element, text)
 	} else {
-		nativeInputValueSetter.call(element, text)
+		getNativeValueSetter(element as HTMLInputElement | HTMLTextAreaElement).call(element, text)
 	}
 
 	// Only dispatch shared input event for non-contenteditable (contenteditable has its own)
@@ -194,9 +235,10 @@ export async function inputTextElement(element: HTMLElement, text: string) {
 
 /**
  * @todo browser-use version is very complex and supports menu tags, need to follow up
+ * @private Internal method, subject to change at any time.
  */
 export async function selectOptionElement(selectElement: HTMLSelectElement, optionText: string) {
-	if (!(selectElement instanceof HTMLSelectElement)) {
+	if (!isSelectElement(selectElement)) {
 		throw new Error('Element is not a select element')
 	}
 
@@ -213,11 +255,14 @@ export async function selectOptionElement(selectElement: HTMLSelectElement, opti
 	await waitFor(0.1) // Wait to ensure change event processing completes
 }
 
-interface ScrollableElement extends HTMLElement {
+interface ScrollableElement extends Element {
 	scrollIntoViewIfNeeded?: (centerIfNeeded?: boolean) => void
 }
 
-export async function scrollIntoViewIfNeeded(element: HTMLElement) {
+/**
+ * @private Internal method, subject to change at any time.
+ */
+export async function scrollIntoViewIfNeeded(element: Element) {
 	const el = element as ScrollableElement
 	if (typeof el.scrollIntoViewIfNeeded === 'function') {
 		el.scrollIntoViewIfNeeded()
@@ -229,11 +274,7 @@ export async function scrollIntoViewIfNeeded(element: HTMLElement) {
 	}
 }
 
-export async function scrollVertically(
-	down: boolean,
-	scroll_amount: number,
-	element?: HTMLElement | null
-) {
+export async function scrollVertically(scroll_amount: number, element?: HTMLElement | null) {
 	// Element-specific scrolling if element is provided
 	if (element) {
 		const targetElement = element
@@ -246,7 +287,10 @@ export async function scrollVertically(
 
 		while (currentElement && attempts < 10) {
 			const computedStyle = window.getComputedStyle(currentElement)
-			const hasScrollableY = /(auto|scroll|overlay)/.test(computedStyle.overflowY)
+			const hasScrollableY =
+				/(auto|scroll|overlay)/.test(computedStyle.overflowY) ||
+				(computedStyle.scrollbarWidth && computedStyle.scrollbarWidth !== 'auto') ||
+				(computedStyle.scrollbarGutter && computedStyle.scrollbarGutter !== 'auto')
 			const canScrollVertically = currentElement.scrollHeight > currentElement.clientHeight
 
 			if (hasScrollableY && canScrollVertically) {
@@ -298,9 +342,20 @@ export async function scrollVertically(
 		el.scrollHeight > el.clientHeight &&
 		bigEnough(el)
 
+	// @deprecated Heuristic container search.
+	// Unreliable in multi-panel layouts. Should guide LLMs to use indexed scroll for consistency.
+	// TODO: remove this fallback
+
+	// try to find the nearest scrollable container
+	// document.activeElement is usually body.
+	// After a successful element.focus(), activeElement become the nearest focusable parent
+
 	let el: HTMLElement | null = document.activeElement as HTMLElement | null
 	while (el && !canScroll(el) && el !== document.body) el = el.parentElement
 
+	// Something is wrong if it falls back to global '*' search
+	// TODO: Return error message instead of global '*' search
+
 	el = canScroll(el)
 		? el
 		: Array.from(document.querySelectorAll<HTMLElement>('*')).find(canScroll) ||
@@ -331,6 +386,10 @@ export async function scrollVertically(
 		return `✅ Scrolled page by ${scrolled}px.`
 	} else {
 		// Container scroll
+
+		const warningMsg = `The document is not scrollable. Falling back to container scroll.`
+		console.log(`[PageController] ${warningMsg}`)
+
 		const scrollBefore = el!.scrollTop
 		const scrollMax = el!.scrollHeight - el!.clientHeight
 
@@ -342,26 +401,22 @@ export async function scrollVertically(
 
 		if (Math.abs(scrolled) < 1) {
 			return dy > 0
-				? `⚠️ Already at the bottom of container (${el!.tagName}), cannot scroll down further.`
-				: `⚠️ Already at the top of container (${el!.tagName}), cannot scroll up further.`
+				? `⚠️ ${warningMsg} Already at the bottom of container (${el!.tagName}), cannot scroll down further.`
+				: `⚠️ ${warningMsg} Already at the top of container (${el!.tagName}), cannot scroll up further.`
 		}
 
 		const reachedBottom = dy > 0 && scrollAfter >= scrollMax - 1
 		const reachedTop = dy < 0 && scrollAfter <= 1
 
 		if (reachedBottom)
-			return `✅ Scrolled container (${el!.tagName}) by ${scrolled}px. Reached the bottom.`
+			return `✅ ${warningMsg} Scrolled container (${el!.tagName}) by ${scrolled}px. Reached the bottom.`
 		if (reachedTop)
-			return `✅ Scrolled container (${el!.tagName}) by ${scrolled}px. Reached the top.`
-		return `✅ Scrolled container (${el!.tagName}) by ${scrolled}px.`
+			return `✅ ${warningMsg} Scrolled container (${el!.tagName}) by ${scrolled}px. Reached the top.`
+		return `✅ ${warningMsg} Scrolled container (${el!.tagName}) by ${scrolled}px.`
 	}
 }
 
-export async function scrollHorizontally(
-	right: boolean,
-	scroll_amount: number,
-	element?: HTMLElement | null
-) {
+export async function scrollHorizontally(scroll_amount: number, element?: HTMLElement | null) {
 	// Element-specific scrolling if element is provided
 	if (element) {
 		const targetElement = element
@@ -370,11 +425,14 @@ export async function scrollHorizontally(
 		let scrolledElement: HTMLElement | null = null
 		let scrollDelta = 0
 		let attempts = 0
-		const dx = right ? scroll_amount : -scroll_amount
+		const dx = scroll_amount
 
 		while (currentElement && attempts < 10) {
 			const computedStyle = window.getComputedStyle(currentElement)
-			const hasScrollableX = /(auto|scroll|overlay)/.test(computedStyle.overflowX)
+			const hasScrollableX =
+				/(auto|scroll|overlay)/.test(computedStyle.overflowX) ||
+				(computedStyle.scrollbarWidth && computedStyle.scrollbarWidth !== 'auto') ||
+				(computedStyle.scrollbarGutter && computedStyle.scrollbarGutter !== 'auto')
 			const canScrollHorizontally = currentElement.scrollWidth > currentElement.clientWidth
 
 			if (hasScrollableX && canScrollHorizontally) {
@@ -418,7 +476,8 @@ export async function scrollHorizontally(
 
 	// Page-level scrolling (default or fallback)
 
-	const dx = right ? scroll_amount : -scroll_amount
+	const dx = scroll_amount
+
 	const bigEnough = (el: HTMLElement) => el.clientWidth >= window.innerWidth * 0.5
 	const canScroll = (el: HTMLElement | null) =>
 		el &&
@@ -426,6 +485,9 @@ export async function scrollHorizontally(
 		el.scrollWidth > el.clientWidth &&
 		bigEnough(el)
 
+	// @deprecated Same heuristic container search as scrollVertically.
+	// TODO: Remove once LLMs reliably use indexed scrolling via data-scrollable.
+
 	let el: HTMLElement | null = document.activeElement as HTMLElement | null
 	while (el && !canScroll(el) && el !== document.body) el = el.parentElement
 
@@ -460,6 +522,9 @@ export async function scrollHorizontally(
 		return `✅ Scrolled page horizontally by ${scrolled}px.`
 	} else {
 		// Container scroll
+		const warningMsg = `The document is not scrollable. Falling back to container scroll.`
+		console.log(`[PageController] ${warningMsg}`)
+
 		const scrollBefore = el!.scrollLeft
 		const scrollMax = el!.scrollWidth - el!.clientWidth
 
@@ -471,17 +536,17 @@ export async function scrollHorizontally(
 
 		if (Math.abs(scrolled) < 1) {
 			return dx > 0
-				? `⚠️ Already at the right edge of container (${el!.tagName}), cannot scroll right further.`
-				: `⚠️ Already at the left edge of container (${el!.tagName}), cannot scroll left further.`
+				? `⚠️ ${warningMsg} Already at the right edge of container (${el!.tagName}), cannot scroll right further.`
+				: `⚠️ ${warningMsg} Already at the left edge of container (${el!.tagName}), cannot scroll left further.`
 		}
 
 		const reachedRight = dx > 0 && scrollAfter >= scrollMax - 1
 		const reachedLeft = dx < 0 && scrollAfter <= 1
 
 		if (reachedRight)
-			return `✅ Scrolled container (${el!.tagName}) by ${scrolled}px. Reached the right edge.`
+			return `✅ ${warningMsg} Scrolled container (${el!.tagName}) by ${scrolled}px. Reached the right edge.`
 		if (reachedLeft)
-			return `✅ Scrolled container (${el!.tagName}) by ${scrolled}px. Reached the left edge.`
-		return `✅ Scrolled container (${el!.tagName}) horizontally by ${scrolled}px.`
+			return `✅ ${warningMsg} Scrolled container (${el!.tagName}) by ${scrolled}px. Reached the left edge.`
+		return `✅ ${warningMsg} Scrolled container (${el!.tagName}) horizontally by ${scrolled}px.`
 	}
 }

packages/page-controller/src/dom/dom_tree/index.js

@@ -18,6 +18,7 @@
  * @edit improve `sampleRect`, filter out rects with 0 area
  * @edit exclude aria-hidden elements
  * @edit make sure attributes exist for interactive candidates.
+ * @edit fix "aria-*" attributes check
  */
 
 export default (
@@ -502,11 +503,16 @@ export default (
 		const overflowX = style.overflowX
 		const overflowY = style.overflowY
 
-		// Check scrollable distances
+		// scrollbar-width/scrollbar-gutter are only set on elements designed to scroll;
+		// their presence signals scroll intent even when overflow is hidden (e.g. overflow: auto on :hover)
+		const hasScrollbarSignal =
+			(style.scrollbarWidth && style.scrollbarWidth !== 'auto') ||
+			(style.scrollbarGutter && style.scrollbarGutter !== 'auto')
+
 		const scrollableX = overflowX === 'auto' || overflowX === 'scroll'
 		const scrollableY = overflowY === 'auto' || overflowY === 'scroll'
 
-		if (!scrollableX && !scrollableY) {
+		if (!scrollableX && !scrollableY && !hasScrollbarSignal) {
 			return null // Not scrollable in any direction
 		}
 
@@ -520,11 +526,11 @@ export default (
 			return null // Not scrollable
 		}
 
-		if (!scrollableY && scrollWidth < threshold) {
+		if (!scrollableY && !hasScrollbarSignal && scrollWidth < threshold) {
 			return null // Not scrollable horizontally
 		}
 
-		if (!scrollableX && scrollHeight < threshold) {
+		if (!scrollableX && !hasScrollbarSignal && scrollHeight < threshold) {
 			return null // Not scrollable vertically
 		}
 
@@ -546,6 +552,8 @@ export default (
 			scrollData: scrollData,
 		})
 
+		console.log('scrollData!!!', scrollData)
+
 		return scrollData
 	}
 
@@ -1143,6 +1151,31 @@ export default (
 	 * @param {HTMLElement} element - The element to check.
 	 * @returns {boolean} Whether the element is an interactive candidate.
 	 */
+
+	// @edit fix "aria-*" attributes check
+	const INTERACTIVE_ARIA_ATTRS = [
+		'aria-expanded',
+		'aria-checked',
+		'aria-selected',
+		'aria-pressed',
+		'aria-haspopup',
+		'aria-controls',
+		'aria-owns',
+		'aria-activedescendant',
+		'aria-valuenow',
+		'aria-valuetext',
+		'aria-valuemax',
+		'aria-valuemin',
+		'aria-autocomplete',
+	]
+
+	function hasInteractiveAria(el) {
+		for (let i = 0; i < INTERACTIVE_ARIA_ATTRS.length; i++) {
+			if (el.hasAttribute(INTERACTIVE_ARIA_ATTRS[i])) return true
+		}
+		return false
+	}
+
 	function isInteractiveCandidate(element) {
 		if (!element || element.nodeType !== Node.ELEMENT_NODE) return false
 
@@ -1167,7 +1200,7 @@ export default (
 			element.hasAttribute('onclick') ||
 			element.hasAttribute('role') ||
 			element.hasAttribute('tabindex') ||
-			element.hasAttribute('aria-') ||
+			hasInteractiveAria(element) ||
 			element.hasAttribute('data-action') ||
 			element.getAttribute('contenteditable') === 'true'
 
@@ -1185,8 +1218,9 @@ export default (
 		'details',
 		'label',
 		'option',
+		'li',
 	])
-	const INTERACTIVE_ROLES = new Set([
+	const DISTINCT_INTERACTIVE_ROLES = new Set([
 		'button',
 		'link',
 		'menuitem',
@@ -1202,6 +1236,9 @@ export default (
 		'searchbox',
 		'textbox',
 		'listbox',
+		'listitem',
+		'treeitem',
+		'row',
 		'option',
 		'scrollbar',
 	])
@@ -1278,7 +1315,7 @@ export default (
 			return true
 		}
 		// Check interactive roles
-		if (role && INTERACTIVE_ROLES.has(role)) {
+		if (role && DISTINCT_INTERACTIVE_ROLES.has(role)) {
 			return true
 		}
 		// Check contenteditable
@@ -1297,6 +1334,10 @@ export default (
 		if (element.hasAttribute('onclick') || typeof element.onclick === 'function') {
 			return true
 		}
+		// ARIA state attributes imply the element manages its own interaction state
+		if (hasInteractiveAria(element)) {
+			return true
+		}
 
 		// return false
 
@@ -1352,6 +1393,12 @@ export default (
 			return true
 		}
 
+		// Scrollable containers are always distinct — the LLM needs their index for targeted scrolling.
+		// Check extraData (already set by isScrollableElement in isInteractiveElement) to avoid redundant layout reads.
+		if (extraData.get(element)?.scrollable) {
+			return true
+		}
+
 		// Default to false: if it's interactive but doesn't match above,
 		// assume it triggers the same action as the parent.
 		return false

packages/page-controller/src/dom/index.ts

@@ -28,8 +28,27 @@ export interface DomConfig {
 	includeAttributes?: string[]
 	highlightOpacity?: number
 	highlightLabelOpacity?: number
+
+	/**
+	 * Preserve semantic landmark tags in dehydrated output even if not interactive
+	 * @note maybe confusing for LLM combining with page scrolling, use with caution
+	 **/
+	keepSemanticTags?: boolean
 }
 
+// TODO: corresponding roles
+const SEMANTIC_TAGS = new Set([
+	'nav',
+	'menu',
+	// 'main',
+	'header',
+	'footer',
+	'aside',
+	// 'article',
+	// 'form',
+	'dialog',
+])
+
 /**
  * 用于检测可交互元素是否是新出现的。
  */
@@ -171,7 +190,11 @@ interface TreeNode {
  *
  * @todo 数据脱敏过滤器
  */
-export function flatTreeToString(flatTree: FlatDomTree, includeAttributes?: string[]): string {
+export function flatTreeToString(
+	flatTree: FlatDomTree,
+	includeAttributes: string[] = [],
+	keepSemanticTags = false
+): string {
 	const DEFAULT_INCLUDE_ATTRIBUTES = [
 		'title',
 		'type',
@@ -203,7 +226,7 @@ export function flatTreeToString(flatTree: FlatDomTree, includeAttributes?: stri
 		'contenteditable',
 	]
 
-	const includeAttrs = [...(includeAttributes || []), ...DEFAULT_INCLUDE_ATTRIBUTES]
+	const includeAttrs = [...includeAttributes, ...DEFAULT_INCLUDE_ATTRIBUTES]
 
 	// Helper function to cap text length
 	const capTextLength = (text: string, maxLength: number): string => {
@@ -294,6 +317,8 @@ export function flatTreeToString(flatTree: FlatDomTree, includeAttributes?: stri
 		const depthStr = '\t'.repeat(depth)
 
 		if (node.type === 'element') {
+			const isSemantic = keepSemanticTags && node.tagName && SEMANTIC_TAGS.has(node.tagName)
+
 			// Add element with highlight_index
 			if (node.highlightIndex !== undefined) {
 				nextDepth += 1
@@ -391,10 +416,30 @@ export function flatTreeToString(flatTree: FlatDomTree, includeAttributes?: stri
 				result.push(line)
 			}
 
-			// Process children regardless
+			// special treatment for semantic tags
+			// even if they are not interactive, we can keep them for clear context
+
+			const emitSemantic = isSemantic && node.highlightIndex === undefined
+			// to check if this tag is empty
+			const mark = emitSemantic ? result.length : -1
+
+			if (emitSemantic) {
+				result.push(`${depthStr}<${node.tagName}>`)
+				nextDepth += 1
+			}
+
 			for (const child of node.children) {
 				processNode(child, nextDepth, result)
 			}
+
+			if (emitSemantic) {
+				// empty tag should be removed
+				if (result.length === mark + 1) {
+					result.pop()
+				} else {
+					result.push(`${depthStr}</${node.tagName}>`)
+				}
+			}
 		} else if (node.type === 'text') {
 			// Add text only if it doesn't have a highlighted parent
 			if (hasParentWithHighlightIndex(node)) {

packages/page-controller/src/mask/SimulatorMask.ts

@@ -5,7 +5,7 @@ import { isPageDark } from './checkDarkMode'
 import styles from './SimulatorMask.module.css'
 import cursorStyles from './cursor.module.css'
 
-export class SimulatorMask {
+export class SimulatorMask extends EventTarget {
 	shown: boolean = false
 	wrapper = document.createElement('div')
 	motion: Motion | null = null
@@ -19,6 +19,8 @@ export class SimulatorMask {
 	#targetCursorY = 0
 
 	constructor() {
+		super()
+
 		this.wrapper.id = 'page-agent-runtime_simulator-mask'
 		this.wrapper.className = styles.wrapper
 		this.wrapper.setAttribute('data-browser-use-ignore', 'true')
@@ -74,13 +76,34 @@ export class SimulatorMask {
 
 		this.#moveCursorToTarget()
 
-		window.addEventListener('PageAgent::MovePointerTo', (event: Event) => {
+		// global events
+		// @note Mask should be isolated from the rest of the code.
+		// Global events are easier to manage and cleanup.
+
+		const movePointerToListener = (event: Event) => {
 			const { x, y } = (event as CustomEvent).detail
 			this.setCursorPosition(x, y)
-		})
-
-		window.addEventListener('PageAgent::ClickPointer', (event: Event) => {
+		}
+		const clickPointerListener = () => {
 			this.triggerClickAnimation()
+		}
+		const enablePassThroughListener = () => {
+			this.wrapper.style.pointerEvents = 'none'
+		}
+		const disablePassThroughListener = () => {
+			this.wrapper.style.pointerEvents = 'auto'
+		}
+
+		window.addEventListener('PageAgent::MovePointerTo', movePointerToListener)
+		window.addEventListener('PageAgent::ClickPointer', clickPointerListener)
+		window.addEventListener('PageAgent::EnablePassThrough', enablePassThroughListener)
+		window.addEventListener('PageAgent::DisablePassThrough', disablePassThroughListener)
+
+		this.addEventListener('dispose', () => {
+			window.removeEventListener('PageAgent::MovePointerTo', movePointerToListener)
+			window.removeEventListener('PageAgent::ClickPointer', clickPointerListener)
+			window.removeEventListener('PageAgent::EnablePassThrough', enablePassThroughListener)
+			window.removeEventListener('PageAgent::DisablePassThrough', disablePassThroughListener)
 		})
 	}
 
@@ -177,7 +200,9 @@ export class SimulatorMask {
 	}
 
 	dispose() {
+		console.log('dispose SimulatorMask')
 		this.motion?.dispose()
 		this.wrapper.remove()
+		this.dispatchEvent(new Event('dispose'))
 	}
 }

packages/page-controller/src/utils/index.ts

@@ -0,0 +1,80 @@
+// ======= type guards =======
+// @note instanceof fails for elements inside iframes
+
+export function isHTMLElement(el: unknown): el is HTMLElement {
+	// @todo either specify to HTMLElement or allow Element here.
+	return !!el && (el as Node).nodeType === 1
+}
+
+export function isInputElement(el: Element): el is HTMLInputElement {
+	return el?.nodeType === 1 && el.tagName === 'INPUT'
+}
+
+export function isTextAreaElement(el: Element): el is HTMLTextAreaElement {
+	return el?.nodeType === 1 && el.tagName === 'TEXTAREA'
+}
+
+export function isSelectElement(el: Element): el is HTMLSelectElement {
+	return el?.nodeType === 1 && el.tagName === 'SELECT'
+}
+
+export function isAnchorElement(el: Element): el is HTMLAnchorElement {
+	return el?.nodeType === 1 && el.tagName === 'A'
+}
+
+// ======= iframe helpers =======
+
+/** Iframe offset for translating element coordinates to top-frame viewport. */
+export function getIframeOffset(element: HTMLElement): { x: number; y: number } {
+	const frame = element.ownerDocument.defaultView?.frameElement as HTMLElement | null
+	if (!frame) return { x: 0, y: 0 }
+	const rect = frame.getBoundingClientRect()
+	return { x: rect.left, y: rect.top }
+}
+
+/**
+ * Get native value setter from the element's own prototype (iframe-safe).
+ * @note for React
+ */
+export function getNativeValueSetter(element: HTMLInputElement | HTMLTextAreaElement) {
+	// eslint-disable-next-line @typescript-eslint/unbound-method
+	return Object.getOwnPropertyDescriptor(Object.getPrototypeOf(element) as object, 'value')!
+		.set as (v: string) => void
+}
+
+// ======= general utils =======
+
+export async function waitFor(seconds: number): Promise<void> {
+	await new Promise((resolve) => setTimeout(resolve, seconds * 1000))
+}
+
+// ======= mask events =======
+
+/**
+ * Move the visual pointer to a position within an element.
+ * @param x - x coordinate in the element's document viewport
+ * @param y - y coordinate in the element's document viewport
+ */
+export async function movePointerToElement(element: HTMLElement, x: number, y: number) {
+	const offset = getIframeOffset(element)
+
+	window.dispatchEvent(
+		new CustomEvent('PageAgent::MovePointerTo', {
+			detail: { x: x + offset.x, y: y + offset.y },
+		})
+	)
+
+	await waitFor(0.3)
+}
+
+export async function clickPointer() {
+	window.dispatchEvent(new CustomEvent('PageAgent::ClickPointer'))
+}
+
+export async function enablePassThrough() {
+	window.dispatchEvent(new CustomEvent('PageAgent::EnablePassThrough'))
+}
+
+export async function disablePassThrough() {
+	window.dispatchEvent(new CustomEvent('PageAgent::DisablePassThrough'))
+}

packages/ui/package.json

@@ -1,6 +1,6 @@
 {
     "name": "@page-agent/ui",
-    "version": "1.5.9",
+    "version": "1.7.1",
     "type": "module",
     "main": "./dist/lib/page-agent-ui.js",
     "module": "./dist/lib/page-agent-ui.js",

packages/ui/src/panel/Panel.ts

@@ -369,6 +369,7 @@ export class Panel {
 	}
 
 	#createWrapper(): HTMLElement {
+		const taskInputMaxLength = 1000
 		const wrapper = document.createElement('div')
 		wrapper.id = 'page-agent-runtime_agent-panel'
 		wrapper.className = styles.wrapper
@@ -406,7 +407,7 @@ export class Panel {
 					<input 
 						type="text" 
 						class="${styles.taskInput}" 
-						maxlength="200"
+						maxlength="${taskInputMaxLength}"
 					/>
 				</div>
 			</div>

packages/website/AGENTS.md

@@ -85,7 +85,7 @@ Uses wouter browser routing with base path for GitHub Pages deployment at `https
 ```tsx
 // main.tsx
 <Router base="/page-agent">
-  <PagesRouter />
+    <PagesRouter />
 </Router>
 ```
 
@@ -109,6 +109,7 @@ Instead of `404.html` redirects, the build copies `index.html` into every route
 2. Add route in `src/pages/docs/index.tsx`
 3. Add navigation item in `src/pages/docs/Layout.tsx`
 4. Add path to `SPA_ROUTES` in `vite.config.js`
+5. Keep `<slug>` consistent across the folder name, import path, route path, sidebar link, and any related cross-links. If the public route changes, rename the folder too.
 
 ## Configuration Files
 

packages/website/package.json

@@ -1,7 +1,7 @@
 {
     "name": "@page-agent/website",
     "private": true,
-    "version": "1.5.9",
+    "version": "1.7.1",
     "type": "module",
     "scripts": {
         "dev": "vite --host 0.0.0.0",
@@ -19,13 +19,13 @@
         "@types/react-dom": "^19.2.1",
         "class-variance-authority": "^0.7.1",
         "clsx": "^2.1.1",
-        "lucide-react": "^0.577.0",
-        "motion": "^12.37.0",
+        "lucide-react": "^1.7.0",
+        "motion": "^12.38.0",
         "next-themes": "^0.4.6",
         "react": "^19.2.4",
         "react-dom": "^19.2.4",
         "rough-notation": "^0.5.1",
-        "simple-icons": "^16.12.0",
+        "simple-icons": "^16.14.0",
         "sonner": "^2.0.7",
         "tailwind-merge": "^3.5.0",
         "tailwindcss": "^4.1.14",

packages/website/src/components/Header.tsx

@@ -18,7 +18,7 @@ export default function Header() {
 	return (
 		<>
 			<header
-				className="relative z-50 bg-white/80 dark:bg-gray-900/80 backdrop-blur-md border-b border-gray-200 dark:border-gray-700"
+				className="sticky top-0 z-50 bg-white/80 dark:bg-gray-900/80 backdrop-blur-md border-b border-gray-200 dark:border-gray-700"
 				role="banner"
 			>
 				<div className="max-w-7xl mx-auto px-6 py-4">

packages/website/src/components/LanguageSwitcher.tsx

@@ -8,8 +8,8 @@ export default function LanguageSwitcher() {
 	const dropdownRef = useRef<HTMLDivElement>(null)
 
 	const languages = [
-		{ code: 'zh-CN' as const, label: '中文' },
 		{ code: 'en-US' as const, label: 'English' },
+		{ code: 'zh-CN' as const, label: '中文' },
 	]
 
 	const currentLanguage = languages.find((lang) => lang.code === language) || languages[0]

packages/website/src/constants.ts

@@ -1,10 +1,10 @@
 // Demo build (auto-init with demo LLM, for quick testing)
 export const CDN_DEMO_URL =
-	'https://cdn.jsdelivr.net/npm/page-agent@1.5.9/dist/iife/page-agent.demo.js'
+	'https://cdn.jsdelivr.net/npm/page-agent@1.7.1/dist/iife/page-agent.demo.js'
 export const CDN_DEMO_CN_URL =
-	'https://registry.npmmirror.com/page-agent/1.5.9/files/dist/iife/page-agent.demo.js'
+	'https://registry.npmmirror.com/page-agent/1.7.1/files/dist/iife/page-agent.demo.js'
 
 // Demo LLM for website testing (homepage quick trial uses flash)
 export const DEMO_MODEL = 'qwen3.5-flash'
 export const DEMO_BASE_URL = 'https://page-ag-testing-ohftxirgbn.cn-shanghai.fcapp.run'
-export const DEMO_API_KEY = 'NA'
+// export const DEMO_API_KEY = ''

packages/website/src/pages/docs/Layout.tsx

@@ -45,6 +45,7 @@ export default function DocsLayout({ children }: DocsLayoutProps) {
 				{ title: isZh ? '知识注入' : 'Instructions', path: '/features/custom-instructions' },
 				{ title: isZh ? '数据脱敏' : 'Data Masking', path: '/features/data-masking' },
 				{ title: isZh ? 'Chrome 扩展' : 'Chrome Extension', path: '/features/chrome-extension' },
+				{ title: 'MCP Server (Beta)', path: '/features/mcp-server' },
 				{
 					title: isZh ? '接入第三方 Agent' : 'Third-party Agent',
 					path: '/features/third-party-agent',

packages/website/src/pages/docs/advanced/page-agent-core/page.tsx

@@ -129,12 +129,6 @@ const result = await agent.execute('Fill in the form with test data')`}
 								? 'LLM API 的基础 URL（如 https://api.openai.com/v1）'
 								: 'Base URL of the LLM API (e.g., https://api.openai.com/v1)',
 						},
-						{
-							name: 'apiKey',
-							type: 'string',
-							required: true,
-							description: isZh ? 'API 密钥' : 'API key for authentication',
-						},
 						{
 							name: 'model',
 							type: 'string',
@@ -143,6 +137,12 @@ const result = await agent.execute('Fill in the form with test data')`}
 								? '模型名称（如 gpt-5.2, anthropic/claude-4.5-haiku）'
 								: 'Model name (e.g., gpt-5.2, anthropic/claude-4.5-haiku)',
 						},
+						{
+							name: 'apiKey',
+							type: 'string',
+							required: false,
+							description: 'LLM AK',
+						},
 						{
 							name: 'temperature',
 							type: 'number',
@@ -156,6 +156,14 @@ const result = await agent.execute('Fill in the form with test data')`}
 							defaultValue: '3',
 							description: isZh ? 'API 调用失败时的最大重试次数' : 'Maximum retries on API failure',
 						},
+						{
+							name: 'disableNamedToolChoice',
+							type: 'boolean',
+							defaultValue: 'false',
+							description: isZh
+								? '禁用命名 tool_choice，始终使用 "required" 字符串。适用于不支持 tool_choice 对象格式的 LLM 服务。'
+								: 'Disable named tool_choice, always use "required" string. For LLM services that don\'t support the object format of tool_choice.',
+						},
 						{
 							name: 'customFetch',
 							type: 'typeof fetch',

packages/website/src/pages/docs/advanced/page-agent/page.tsx

@@ -100,7 +100,7 @@ console.log(result.history) // Full execution history`}
 							>
 								AgentConfig
 							</Link>{' '}
-							和{' '}
+							、 PanelConfig 和{' '}
 							<Link
 								href="/advanced/page-controller#configuration"
 								className="text-blue-600 dark:text-blue-400 hover:underline"
@@ -125,7 +125,7 @@ console.log(result.history) // Full execution history`}
 							>
 								AgentConfig
 							</Link>{' '}
-							and{' '}
+							, PanelConfig, and{' '}
 							<Link
 								href="/advanced/page-controller#configuration"
 								className="text-blue-600 dark:text-blue-400 hover:underline"

packages/website/src/pages/docs/advanced/page-controller/page.tsx

@@ -108,6 +108,14 @@ const agent = new PageAgentCore({
 								? '在 DOM 提取中包含的额外 HTML 属性。支持通配符 *（如 data-* 匹配所有 data- 开头的属性）。默认已包含常见属性如 role, aria-label 等。'
 								: 'Additional HTML attributes to include in DOM extraction. Supports wildcard * (e.g. data-* matches all data- prefixed attributes). Common attributes like role, aria-label are included by default.',
 						},
+						{
+							name: 'keepSemanticTags',
+							type: 'boolean',
+							defaultValue: 'false',
+							description: isZh
+								? '在简化输出中保留语义标签（如 nav, main, header, footer, aside 等），即使它们不可交互。帮助 LLM 理解页面结构。'
+								: 'Preserve semantic landmark tags (e.g. nav, main, header, footer, aside) in dehydrated output even if not interactive. Helps LLM understand page structure.',
+						},
 					]}
 				/>
 			</section>

packages/website/src/pages/docs/features/chrome-extension/page.tsx

@@ -187,39 +187,7 @@ localStorage.setItem('PageAgentExtUserAuthToken', '<your-token-from-extension>')
 					</p>
 
 					<CodeEditor
-						code={
-							isZh
-								? `import type {
-	AgentActivity,
-	AgentStatus,
-	ExecutionResult,
-	HistoricalEvent
-} from '@page-agent/core'
-
-interface ExecuteConfig {
-	baseURL: string   // LLM API 端点
-	apiKey: string    // API 密钥
-	model: string     // 模型名称
-
-	includeInitialTab?: boolean
-	onStatusChange?: (status: AgentStatus) => void
-	onActivity?: (activity: AgentActivity) => void
-	onHistoryUpdate?: (history: HistoricalEvent[]) => void
-}
-
-type Execute = (task: string, config: ExecuteConfig) => Promise<ExecutionResult>
-
-declare global {
-	interface Window {
-		PAGE_AGENT_EXT_VERSION?: string
-		PAGE_AGENT_EXT?: {
-			version: string
-			execute: Execute
-			stop: () => void
-		}
-	}
-}`
-								: `import type {
+						code={`import type {
 	AgentActivity,
 	AgentStatus,
 	ExecutionResult,
@@ -228,10 +196,12 @@ declare global {
 
 interface ExecuteConfig {
 	baseURL: string   // LLM API endpoint
-	apiKey: string    // API key
 	model: string     // Model name
+	apiKey?: string   // LLM AK
 
+	systemInstruction?: string // Global system-level instructions
 	includeInitialTab?: boolean
+	experimentalIncludeAllTabs?: boolean // Control all unpinned tabs in the window
 	onStatusChange?: (status: AgentStatus) => void
 	onActivity?: (activity: AgentActivity) => void
 	onHistoryUpdate?: (history: HistoricalEvent[]) => void
@@ -248,8 +218,7 @@ declare global {
 			stop: () => void
 		}
 	}
-}`
-						}
+}`}
 						language="typescript"
 					/>
 
@@ -266,6 +235,7 @@ const result = await window.PAGE_AGENT_EXT.execute(
 		apiKey: 'your-api-key',
 		model: 'gpt-5.2',
 		// includeInitialTab: false, // 设为 false 排除初始标签页
+		// experimentalIncludeAllTabs: true, // 控制窗口内所有非固定标签页
 		onStatusChange: status => console.log('状态变化:', status),
 		onActivity: activity => console.log('活动:', activity),
 		onHistoryUpdate: history => console.log('历史更新:', history)
@@ -281,6 +251,7 @@ const result = await window.PAGE_AGENT_EXT.execute(
 		apiKey: 'your-api-key',
 		model: 'gpt-5.2',
 		// includeInitialTab: false, // Set to false to exclude initial tab
+		// experimentalIncludeAllTabs: true, // Control all unpinned tabs in the window
 		onStatusChange: status => console.log('Status change:', status),
 		onActivity: activity => console.log('Activity:', activity),
 		onHistoryUpdate: history => console.log('History update:', history)

packages/website/src/pages/docs/features/mcp-server/page.tsx

@@ -0,0 +1,70 @@
+import BetaNotice from '@/components/BetaNotice'
+import CodeEditor from '@/components/CodeEditor'
+import { Heading } from '@/components/Heading'
+
+export default function McpServerPage() {
+	return (
+		<div>
+			<h1 className="text-4xl font-bold mb-6">MCP Server (Beta)</h1>
+			<BetaNotice />
+			<p className="text-xl text-gray-600 dark:text-gray-300 mb-8 leading-relaxed">
+				Use the MCP server to let your local agent send natural-language browser tasks to Page Agent
+				Ext.
+			</p>
+
+			<section className="mb-10">
+				<Heading id="quick-start" className="text-2xl font-bold mb-4">
+					How to use
+				</Heading>
+				<div className="space-y-4">
+					<div className="p-4 bg-blue-50 dark:bg-blue-950/20 rounded-lg border border-blue-200 dark:border-blue-800">
+						<p className="text-sm text-blue-900 dark:text-blue-200 leading-7">
+							1. Install Page Agent Ext in Chrome.
+							<br />
+							2. Add the MCP server to your local agent client.
+							<br />
+							3. Start the client and approve the Hub connection in the browser when prompted.
+							<br />
+							4. Ask your agent to do something in the browser. The client will call execute_task
+							for you.
+						</p>
+					</div>
+
+					<CodeEditor
+						code={`{
+  "mcpServers": {
+    "page-agent": {
+      "command": "npx",
+      "args": ["-y", "@page-agent/mcp"],
+      "env": {
+        "LLM_BASE_URL": "https://api.openai.com/v1",
+        "LLM_API_KEY": "sk-xxx",
+        "LLM_MODEL_NAME": "gpt-5.2"
+      }
+    }
+  }
+}`}
+						language="json"
+					/>
+				</div>
+			</section>
+
+			<section className="mb-10">
+				<Heading id="the-hub" className="text-2xl font-bold mb-4">
+					The Hub
+				</Heading>
+
+				<p className="text-gray-700 dark:text-gray-300 leading-relaxed">
+					The Hub is the control center for communication between Page Agent Ext and external
+					callers.
+				</p>
+				<p className="text-gray-700 dark:text-gray-300 leading-relaxed">
+					When the MCP server starts, it opens a local launcher page. The launcher asks the
+					extension to open the Hub tab, and the Hub receives tasks from your local agent. MCP uses
+					this path, but the Hub itself is the extension's general external communication entry
+					point.
+				</p>
+			</section>
+		</div>
+	)
+}

packages/website/src/pages/docs/features/models/page.tsx

@@ -16,12 +16,12 @@ const BASELINE = new Set([
 // Models grouped by brand, newest first
 const MODEL_GROUPS: Record<string, string[]> = {
 	Qwen: [
+		'qwen3.6-plus',
 		'qwen3.5-plus',
 		'qwen3.5-flash',
 		'qwen3-coder-next',
 		'qwen-3-max',
 		'qwen-3-plus',
-		'qwen3:14b (ollama)',
 	],
 	OpenAI: ['gpt-5.4', 'gpt-5.2', 'gpt-5.1', 'gpt-5', 'gpt-5-mini', 'gpt-4.1', 'gpt-4.1-mini'],
 	DeepSeek: ['deepseek-3.2'],
@@ -33,8 +33,8 @@ const MODEL_GROUPS: Record<string, string[]> = {
 		'claude-haiku-4.5',
 		'claude-sonnet-3.5',
 	],
+	MiniMax: ['MiniMax-M2.7', 'MiniMax-M2.7-highspeed', 'MiniMax-M2.5', 'MiniMax-M2.5-highspeed'],
 	xAI: ['grok-4.1-fast', 'grok-4', 'grok-code-fast'],
-	MiniMax: ['MiniMax-M2.5', 'MiniMax-M2.5-highspeed'],
 	MoonshotAI: ['kimi-k2.5'],
 	'Z.AI': ['glm-5', 'glm-4.7'],
 }
@@ -64,7 +64,6 @@ export default function Models() {
 					: 'Supports models that comply with OpenAI API specification and support tool calls, including public cloud services and private deployments.'}
 			</p>
 
-			{/* Models Section */}
 			<section className="mb-10">
 				<Heading id="tested-models" className="text-2xl font-semibold mb-3">
 					{isZh ? '已测试模型' : 'Tested Models'}
@@ -87,9 +86,10 @@ export default function Models() {
 				</div>
 			</section>
 
-			{/* Tips Section */}
 			<section className="mb-10">
-				<h2 className="text-2xl font-semibold mb-4">{isZh ? '提示' : 'Tips'}</h2>
+				<Heading id="tips" level={3}>
+					Tips
+				</Heading>
 				<div className="p-4 bg-blue-50 dark:bg-blue-950/20 rounded-lg border border-blue-200 dark:border-blue-800">
 					<ul className="text-sm text-gray-700 dark:text-gray-300 space-y-2 list-disc pl-5">
 						<li>
@@ -111,7 +111,6 @@ export default function Models() {
 				</div>
 			</section>
 
-			{/* Configuration Section */}
 			<section className="mb-10">
 				<Heading id="configuration">{isZh ? '配置方式' : 'Configuration'}</Heading>
 				<CodeEditor
@@ -121,26 +120,50 @@ const pageAgent = new PageAgent({
   apiKey: 'your-api-key',
   model: 'qwen3.5-plus'
 });
-
-// MiniMax
-const pageAgent = new PageAgent({
-  baseURL: 'https://api.minimax.io/v1',
-  apiKey: 'your-minimax-api-key',
-  model: 'MiniMax-M2.5'
-});
-
-// Self-hosted models (e.g., Ollama)
-const pageAgent = new PageAgent({
-  baseURL: 'http://localhost:11434/v1',
-  apiKey: 'NA',
-  model: 'qwen3:14b'
-});
-
 `}
 				/>
 			</section>
 
-			{/* Free Testing API Section */}
+			<section className="mb-10">
+				<Heading id="production-authentication" className="text-2xl font-semibold mb-4">
+					{isZh ? '🔐 生产环境鉴权' : '🔐 Production Authentication'}
+				</Heading>
+				<p className="text-sm text-gray-600 dark:text-gray-400 mb-3">
+					{isZh
+						? '如果你只是将它用作个人助手，可以直接连接你的 LLM 服务。'
+						: 'If you only use it as a personal assistant, you can connect to your LLM service directly.'}
+				</p>
+				<p className="text-sm text-gray-600 dark:text-gray-400 mb-3">
+					{isZh ? (
+						<>
+							如果你计划将它集成到你的 Web 应用中，建议搭建一个后端代理来转发 LLM 请求，并使用{' '}
+							<code>customFetch</code> 携带 Cookie 或其他鉴权信息：
+						</>
+					) : (
+						<>
+							If you plan to integrate it into your web app, it's better to have a backend proxy for
+							the LLM and use <code>customFetch</code> to authenticate the request with cookies or
+							other methods:
+						</>
+					)}
+				</p>
+				<CodeEditor
+					code={`const agent = new PageAgent({
+  baseURL: '/api/llm-proxy',
+  model: 'gpt-5.1',
+  customFetch: (url, init) =>
+    fetch(url, { ...init, credentials: 'include' }),
+});`}
+				/>
+				<div className="mt-4 bg-yellow-50 dark:bg-yellow-950/20 border-l-4 border-yellow-500 p-4 rounded-r-lg">
+					<p className="text-sm font-semibold text-yellow-900 dark:text-yellow-200">
+						{isZh
+							? '⚠️ 永远不要把真实的 LLM API Key 提交到前端代码中'
+							: '⚠️ NEVER commit real LLM API keys to your frontend code'}
+					</p>
+				</div>
+			</section>
+
 			<section className="mb-10">
 				<Heading id="free-testing-api">{isZh ? '免费测试接口' : 'Free Testing API'}</Heading>
 				<p className="text-sm text-gray-600 dark:text-gray-400 mb-4">
@@ -182,116 +205,143 @@ const pageAgent = new PageAgent({
 						</a>
 					</p>
 					<CodeEditor
-						code={`# qwen3.5-plus (default for demos) or qwen3.5-flash (lighter)
+						code={`# qwen3.5-plus / qwen3.5-flash
 LLM_BASE_URL="https://page-ag-testing-ohftxirgbn.cn-shanghai.fcapp.run"
-LLM_MODEL_NAME="qwen3.5-plus"
-LLM_API_KEY="NA"`}
+LLM_MODEL_NAME="qwen3.5-plus"`}
 					/>
 				</div>
 			</section>
 
-			{/* Ollama Section */}
 			<section className="mb-10">
-				<Heading id="ollama">Ollama</Heading>
-				<p className="text-sm text-gray-600 dark:text-gray-400 mb-4">
+				<Heading id="local-runtimes">{isZh ? '本地运行时' : 'Local Runtimes'}</Heading>
+				<p className="text-sm text-gray-600 dark:text-gray-400 mb-6">
 					{isZh
-						? '已在 Ollama 0.15 + qwen3:14b (RTX3090 24GB) 上测试通过。'
-						: 'Tested on Ollama 0.15 with qwen3:14b (RTX3090 24GB).'}
+						? '通过 Ollama、LM Studio 等本地 OpenAI-compatible 运行时接入 PageAgent，实现离线或局域网部署。'
+						: 'Use local OpenAI-compatible runtimes such as Ollama and LM Studio with PageAgent for offline or LAN deployments.'}
 				</p>
-				<CodeEditor
-					code={`LLM_BASE_URL="http://localhost:11434/v1"
-LLM_API_KEY="NA"
-LLM_MODEL_NAME="qwen3:14b"`}
-				/>
-				<div className="mt-4 p-4 bg-amber-50 dark:bg-amber-950/20 rounded-lg border border-amber-200 dark:border-amber-800">
-					<h3 className="font-semibold text-amber-900 dark:text-amber-200 mb-2">
-						{isZh ? '⚠️ 注意事项' : '⚠️ Important Notes'}
-					</h3>
-					<ul className="text-sm text-gray-700 dark:text-gray-300 space-y-2 list-disc pl-5">
-						<li>
-							{isZh
-								? '确保 OLLAMA_ORIGINS 设置为 * 以避免 403 错误'
-								: 'Add * to OLLAMA_ORIGINS to avoid 403 errors'}
-						</li>
-						<li>
-							{isZh
-								? '小于 10B 参数的模型通常效果不佳'
-								: 'Models smaller than 10B are unlikely to be strong enough'}
-						</li>
-						<li>{isZh ? '需要支持 tool_call 的模型' : 'Requires tool_call capable models'}</li>
-						<li>
-							{isZh
-								? '确保上下文长度大于输入 token 数，否则 Ollama 会静默截断 prompt。普通页面约需 15k token，随步骤增加。默认 4k 上下文长度无法正常工作'
-								: 'Ensure context length exceeds input tokens, or Ollama will silently truncate prompts. ~15k tokens for a typical page, increases with steps. Default 4k context length will NOT work'}
-						</li>
-					</ul>
-				</div>
 
-				<div className="mt-4">
-					<h3 className="font-semibold text-gray-900 dark:text-gray-100 mb-3">
-						{isZh ? '建议启动参数' : 'Recommended Startup'}
-					</h3>
-					<p className="text-sm text-gray-600 dark:text-gray-400 mb-3">
-						{isZh
-							? '启动 Ollama 时建议配置以下环境变量：扩大上下文窗口、允许跨域访问、监听所有网络接口。'
-							: 'Start Ollama with these environment variables: larger context window, allow cross-origin access, and listen on all interfaces.'}
-					</p>
+				<div className="space-y-10">
+					<section>
+						<Heading id="requirements" level={3}>
+							Requirements
+						</Heading>
+						<div className="mt-4 p-5 bg-amber-50 dark:bg-amber-950/20 rounded-xl border border-amber-200 dark:border-amber-800">
+							<ul className="text-sm text-gray-700 dark:text-gray-300 space-y-2 list-disc pl-5">
+								<li>
+									{isZh
+										? '务必打开 CORS，否则浏览器无法直接请求本地 LLM 服务。'
+										: 'Enable CORS, otherwise the browser cannot call your local LLM endpoint directly.'}
+								</li>
+								<li>
+									{isZh
+										? '将 context length 或 content length 至少设置为 8000。普通页面常常需要 15k token 左右，默认 4k 很容易被截断。'
+										: 'Set context length or content length to at least 8000. A typical page often needs around 15k tokens, so the default 4k usually truncates prompts.'}
+								</li>
+								<li>
+									{isZh ? '需要支持 tool_call 的模型。' : 'Use a model with tool_call support.'}
+								</li>
+								<li>
+									{isZh
+										? '小于 10B 参数的模型通常效果不佳。'
+										: 'Models smaller than 10B are usually not strong enough.'}
+								</li>
+							</ul>
+						</div>
+					</section>
 
-					<div className="space-y-2">
-						<p className="text-xs font-medium text-gray-500 dark:text-gray-400">macOS / Linux</p>
+					<section>
+						<Heading id="local-configuration" level={3}>
+							{isZh ? '基础配置' : 'Basic Configuration'}
+						</Heading>
 						<CodeEditor
-							code={`OLLAMA_CONTEXT_LENGTH=64000 OLLAMA_HOST=0.0.0.0:11434 OLLAMA_ORIGINS="*" ollama serve`}
-						/>
+							code={`// Local OpenAI-compatible runtime - no apiKey needed
+const pageAgent = new PageAgent({
+  baseURL: 'http://localhost:11434/v1',
+  model: 'qwen3:14b'
+});
 
-						<p className="text-xs font-medium text-gray-500 dark:text-gray-400 pt-2">
-							Windows (PowerShell)
+// Or connect to LM Studio
+const lmStudioAgent = new PageAgent({
+  baseURL: 'http://127.0.0.1:1234/v1',
+  model: 'qwen/qwen3.5-27b'
+});
+`}
+						/>
+					</section>
+
+					<section>
+						<Heading id="ollama" level={3}>
+							Ollama
+						</Heading>
+						<p className="text-sm text-gray-600 dark:text-gray-400 mb-4">
+							{isZh
+								? '已在 Ollama 0.15 + qwen3:14b (RTX3090 24GB) 上测试通过。'
+								: 'Tested on Ollama 0.15 with qwen3:14b (RTX3090 24GB).'}
 						</p>
 						<CodeEditor
-							code={`$env:OLLAMA_CONTEXT_LENGTH=64000; $env:OLLAMA_HOST="0.0.0.0:11434"; $env:OLLAMA_ORIGINS="*"; ollama serve`}
+							code={`LLM_BASE_URL="http://localhost:11434/v1"
+LLM_MODEL_NAME="qwen3:14b"`}
 						/>
-					</div>
-				</div>
-			</section>
+						<div className="mt-4 p-4 bg-amber-50 dark:bg-amber-950/20 rounded-lg border border-amber-200 dark:border-amber-800">
+							<h3 className="font-semibold text-amber-900 dark:text-amber-200 mb-2">
+								{isZh ? '⚠️ 注意事项' : '⚠️ Important Notes'}
+							</h3>
+							<p className="text-sm text-gray-700 dark:text-gray-300">
+								{isZh
+									? '如果浏览器侧请求失败，优先检查 Ollama 是否已按上面的要求开启 CORS。'
+									: 'If browser-side requests fail, check whether Ollama has CORS enabled as required above.'}
+							</p>
+						</div>
 
-			{/* Production Authentication */}
-			<section className="mb-10">
-				<Heading id="production-authentication" className="text-2xl font-semibold mb-4">
-					{isZh ? '🔐 生产环境鉴权' : '🔐 Production Authentication'}
-				</Heading>
-				<p className="text-sm text-gray-600 dark:text-gray-400 mb-3">
-					{isZh
-						? '如果你只是将它用作个人助手，可以直接连接你的 LLM 服务。'
-						: 'If you only use it as a personal assistant, you can connect to your LLM service directly.'}
-				</p>
-				<p className="text-sm text-gray-600 dark:text-gray-400 mb-3">
-					{isZh ? (
-						<>
-							如果你计划将它集成到你的 Web 应用中，建议搭建一个后端代理来转发 LLM 请求，并使用{' '}
-							<code>customFetch</code> 携带 Cookie 或其他鉴权信息：
-						</>
-					) : (
-						<>
-							If you plan to integrate it into your web app, it's better to have a backend proxy for
-							the LLM and use <code>customFetch</code> to authenticate the request with cookies or
-							other methods:
-						</>
-					)}
-				</p>
-				<CodeEditor
-					code={`const agent = new PageAgent({
-  baseURL: '/api/llm-proxy',
-  apiKey: 'NA',
-  model: 'gpt-5.1',
-  customFetch: (url, init) =>
-    fetch(url, { ...init, credentials: 'include' }),
-});`}
-				/>
-				<div className="mt-4 bg-yellow-50 dark:bg-yellow-950/20 border-l-4 border-yellow-500 p-4 rounded-r-lg">
-					<p className="text-sm font-semibold text-yellow-900 dark:text-yellow-200">
-						{isZh
-							? '⚠️ 永远不要把真实的 LLM API Key 提交到前端代码中'
-							: '⚠️ NEVER commit real LLM API keys to your frontend code'}
-					</p>
+						<div className="mt-4">
+							<h3 className="font-semibold text-gray-900 dark:text-gray-100 mb-3">
+								{isZh ? '建议启动参数' : 'Recommended Startup'}
+							</h3>
+							<p className="text-sm text-gray-600 dark:text-gray-400 mb-3">
+								{isZh
+									? '启动 Ollama 时建议同时放大上下文窗口并开启跨域访问。'
+									: 'When starting Ollama, increase the context window and enable cross-origin access.'}
+							</p>
+
+							<div className="space-y-2">
+								<p className="text-xs font-medium text-gray-500 dark:text-gray-400">
+									macOS / Linux
+								</p>
+								<CodeEditor
+									code={`OLLAMA_CONTEXT_LENGTH=64000 OLLAMA_HOST=0.0.0.0:11434 OLLAMA_ORIGINS="*" ollama serve`}
+								/>
+
+								<p className="text-xs font-medium text-gray-500 dark:text-gray-400 pt-2">
+									Windows (PowerShell)
+								</p>
+								<CodeEditor
+									code={`$env:OLLAMA_CONTEXT_LENGTH=64000; $env:OLLAMA_HOST="0.0.0.0:11434"; $env:OLLAMA_ORIGINS="*"; ollama serve`}
+								/>
+							</div>
+						</div>
+					</section>
+
+					<section>
+						<Heading id="lm-studio" level={3}>
+							LM Studio
+						</Heading>
+						<CodeEditor
+							code={`LLM_BASE_URL="http://127.0.0.1:1234/v1"
+LLM_MODEL_NAME="qwen/qwen3.5-27b"`}
+						/>
+						<div className="mt-4 p-4 bg-amber-50 dark:bg-amber-950/20 rounded-lg border border-amber-200 dark:border-amber-800">
+							<h3 className="font-semibold text-amber-900 dark:text-amber-200 mb-2">
+								{isZh ? '⚠️ 注意事项' : '⚠️ Important Notes'}
+							</h3>
+							<ul className="text-sm text-gray-700 dark:text-gray-300 space-y-2 list-disc pl-5">
+								<li>
+									{isZh
+										? 'Agent 必须启用 disableNamedToolChoice，否则 tool_choice 参数会报错。'
+										: 'Enable disableNamedToolChoice in the agent config, otherwise the tool_choice parameter may fail.'}
+								</li>
+							</ul>
+						</div>
+					</section>
 				</div>
 			</section>
 		</div>

packages/website/src/pages/docs/index.tsx

@@ -13,6 +13,7 @@ import ChromeExtension from './features/chrome-extension/page'
 import Instructions from './features/custom-instructions/page'
 import CustomTools from './features/custom-tools/page'
 import DataMasking from './features/data-masking/page'
+import McpServerPage from './features/mcp-server/page'
 import Models from './features/models/page'
 import ThirdPartyAgent from './features/third-party-agent/page'
 import Limitations from './introduction/limitations/page'
@@ -80,6 +81,11 @@ export default function DocsRouter() {
 					<ChromeExtension />
 				</DocsPage>
 			</Route>
+			<Route path="/features/mcp-server">
+				<DocsPage>
+					<McpServerPage />
+				</DocsPage>
+			</Route>
 			<Route path="/features/third-party-agent">
 				<DocsPage>
 					<ThirdPartyAgent />

packages/website/src/pages/docs/introduction/limitations/page.tsx

@@ -97,6 +97,7 @@ export default function LimitationsPage() {
 									isZh ? '点击、文本输入、选择' : 'Click, text input, select',
 									isZh ? '页面滚动（垂直 / 水平）' : 'Scroll (vertical / horizontal)',
 									isZh ? '表单提交、焦点切换' : 'Form submit, focus',
+									isZh ? '同源 iframe（仅单层）' : 'Same-origin iframe (single level only)',
 									isZh ? '执行 JavaScript（可选）' : 'Execute JavaScript (opt-in)',
 								].map((text) => (
 									<li key={text} className="flex items-center text-gray-700 dark:text-gray-300">
@@ -115,6 +116,7 @@ export default function LimitationsPage() {
 									isZh ? '悬停、拖拽、右键菜单' : 'Hover, drag & drop, right-click',
 									isZh ? '键盘快捷键' : 'Keyboard shortcuts',
 									isZh ? '坐标定位操作' : 'Position-based control',
+									isZh ? '嵌套 iframe、跨域 iframe' : 'Nested iframes, cross-origin iframes',
 									isZh ? '绘图操作' : 'Drawing',
 									isZh
 										? 'Monaco、CodeMirror 等需要通过 JS 实例控制的编辑器'

packages/website/src/pages/docs/introduction/troubleshooting/page.tsx

@@ -274,10 +274,12 @@ function ApiErrorsContent(isZh: boolean) {
 					code={`const agent = new PageAgent({
   // ...
   customFetch: async (url, init) => {
-    const body = JSON.parse(init.body)
     // Adapt parameters for your provider
-    delete body.stream_options
-    return fetch(url, { ...init, body: JSON.stringify(body) })
+    const body = JSON.parse(init.body)
+    delete body.tool_choice
+    const bodyStr = JSON.stringify(body)
+
+    return fetch(url, { ...init, body: bodyStr })
   },
 })`}
 				/>

packages/website/src/pages/home/FeaturesSection.tsx

@@ -16,12 +16,13 @@ const LLM_CLOUD: {
 	opacity: number
 }[] = [
 	{ name: 'OpenAI', color: '#10b981', x: 18, y: 22, size: 1.5, opacity: 1 },
-	{ name: 'Claude', color: '#f97316', x: 62, y: 15, size: 1.35, opacity: 0.95 },
-	{ name: 'Qwen', color: '#8b5cf6', x: 38, y: 50, size: 1.8, opacity: 0.9 },
-	{ name: 'Gemini', color: '#3b82f6', x: 68, y: 48, size: 1.2, opacity: 0.85 },
-	{ name: 'DeepSeek', color: '#06b6d4', x: 10, y: 65, size: 1.1, opacity: 0.8 },
+	{ name: 'Claude', color: '#f97316', x: 58, y: 15, size: 1.35, opacity: 0.95 },
+	{ name: 'Qwen', color: '#8b5cf6', x: 35, y: 50, size: 1.8, opacity: 0.9 },
+	{ name: 'Gemini', color: '#3b82f6', x: 72, y: 48, size: 1.2, opacity: 0.85 },
+	{ name: 'DeepSeek', color: '#06b6d4', x: 13, y: 65, size: 1.1, opacity: 0.8 },
 	{ name: 'Grok', color: '#f43f5e', x: 52, y: 78, size: 1.0, opacity: 0.75 },
-	{ name: 'Ollama', color: '#9ca3af', x: 82, y: 25, size: 1.1, opacity: 0.8 },
+	{ name: 'Ollama', color: '#9ca3af', x: 84, y: 25, size: 1.1, opacity: 0.8 },
+	{ name: 'LM Studio', color: '#4338ca', x: 49, y: 36, size: 0.8, opacity: 0.9 },
 	{ name: 'Kimi', color: '#14b8a6', x: 30, y: 82, size: 0.85, opacity: 0.6 },
 	{ name: 'GLM', color: '#f59e0b', x: 70, y: 72, size: 0.85, opacity: 0.55 },
 	{ name: 'LLaMA', color: '#60a5fa', x: 88, y: 70, size: 0.8, opacity: 0.45 },

packages/website/src/pages/home/HeroSection.tsx

@@ -10,7 +10,7 @@ import { Particles } from '../../components/ui/particles'
 import {
 	CDN_DEMO_CN_URL,
 	CDN_DEMO_URL,
-	DEMO_API_KEY,
+	// DEMO_API_KEY,
 	DEMO_BASE_URL,
 	DEMO_MODEL,
 } from '../../constants'
@@ -94,7 +94,7 @@ export default function HeroSection() {
 				apiKey:
 					import.meta.env.DEV && import.meta.env.LLM_API_KEY
 						? import.meta.env.LLM_API_KEY
-						: DEMO_API_KEY,
+						: undefined,
 			})
 		}
 
@@ -103,10 +103,7 @@ export default function HeroSection() {
 	}
 
 	return (
-		<section
-			className="relative px-6 pt-24 py-20 pb-18 lg:py-22 lg:pt-28 overflow-hidden"
-			aria-labelledby="hero-heading"
-		>
+		<section className="relative px-6 pt-18 pb-14 lg:pb-20 lg:pt-24" aria-labelledby="hero-heading">
 			<div className="max-w-7xl mx-auto text-center">
 				{/* Background Pattern + Particles */}
 				<div className="absolute inset-0 opacity-30" aria-hidden="true">
@@ -122,7 +119,7 @@ export default function HeroSection() {
 				/>
 
 				<div className="relative z-10">
-					<div className="inline-flex items-center px-4 py-2 mb-8 text-sm font-medium bg-white/90 dark:bg-gray-800/90 rounded-full shadow-lg border border-gray-200 dark:border-gray-700">
+					<div className="inline-flex items-center px-4 py-2 mb-4 text-sm font-medium bg-white/90 dark:bg-gray-800/90 rounded-full shadow-lg border border-gray-200 dark:border-gray-700">
 						<span
 							className="w-2 h-2 bg-blue-500 rounded-full mr-2 animate-pulse"
 							aria-hidden="true"
@@ -134,7 +131,7 @@ export default function HeroSection() {
 
 					<h1
 						id="hero-heading"
-						className="text-5xl lg:text-7xl font-bold mb-14 mt-8 bg-linear-to-r from-blue-600 to-purple-600 bg-clip-text text-transparent pb-1"
+						className="text-5xl lg:text-7xl font-bold mb-10 mt-8 bg-linear-to-r from-blue-600 to-purple-600 bg-clip-text text-transparent pb-1"
 					>
 						{isZh ? (
 							<>

packages/website/src/pages/home/OneMoreThingSection.tsx

@@ -58,6 +58,22 @@ export default function OneMoreThingSection() {
 					</Link>
 				</div>
 
+				<div className="mb-10 rounded-2xl border border-blue-200/70 dark:border-blue-800/70 bg-linear-to-r from-blue-50 to-white dark:from-blue-950/30 dark:to-gray-900 px-5 py-4 max-w-3xl mx-auto text-left sm:text-center">
+					<p className="text-sm text-gray-700 dark:text-gray-300 leading-7">
+						{isZh
+							? '从 Claude Desktop、Copilot 或其他本地 Agent 直接发起浏览器任务？'
+							: 'Using Claude Desktop, Copilot, or another local agent? Connect it to the extension with the MCP server.'}
+					</p>
+					<p>
+						<Link
+							href="/docs/features/mcp-server"
+							className="font-medium text-blue-700 dark:text-blue-300 underline underline-offset-4"
+						>
+							{isZh ? '查看 MCP 文档' : 'Read the MCP docs'}
+						</Link>
+					</p>
+				</div>
+
 				<div className="grid sm:grid-cols-3 gap-5 text-left max-w-3xl mx-auto">
 					{[
 						{
@@ -67,16 +83,16 @@ export default function OneMoreThingSection() {
 								: 'Run tasks across multiple pages and tabs without being limited to a single page context',
 						},
 						{
-							title: isZh ? '页面内发起控制' : 'Control from Your Page',
+							title: isZh ? '从页面发起控制' : 'Control from a WebPage',
 							desc: isZh
 								? '在页面 JS 中发起任务，驱动整个浏览器完成跨标签操作'
-								: 'Trigger tasks from page JS to drive the entire browser across tabs',
+								: 'Trigger tasks from in-page JS to drive the entire browser across tabs',
 						},
 						{
-							title: isZh ? '外部发起任务' : 'External Triggers',
+							title: isZh ? '外部发起任务' : 'External Caller',
 							desc: isZh
 								? '页面 JS、本地 Agent 或云端 Agent 均可通过扩展发起任务'
-								: 'Page JS, local agents, or cloud agents can trigger tasks through the extension',
+								: 'Local agents and cloud agents can control user browser through the extension',
 						},
 					].map((item) => (
 						<MagicCard

packages/website/vite.config.js

@@ -26,7 +26,9 @@ const SPA_ROUTES = [
 	'docs/features/data-masking',
 	'docs/features/custom-instructions',
 	'docs/features/models',
+	'docs/features/local-llms',
 	'docs/features/chrome-extension',
+	'docs/features/mcp-server',
 	'docs/features/third-party-agent',
 	'docs/advanced/page-agent',
 	'docs/advanced/page-agent-core',