Control Chrome tabs, interact with web elements, and automate complex browser workflows using an AI-powered MCP bridge and extension.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install skill-6
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install skill-6 using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is AIPex Browser Control?
AIPex Browser Control is a sophisticated automation bridge that enables AI agents to command a Chrome browser via the Model Context Protocol (MCP). By connecting a local bridge to the AIPex Chrome extension, agents gain the ability to navigate the web, interact with DOM elements, and manage tab lifecycles using natural language. This skill is essential for developers building autonomous agents within the ecosystem of Openclaw Skills that require real-time web interaction.
The architecture relies on a WebSocket connection between the browser extension and an MCP bridge, allowing for low-latency execution of over 30 specialized tools. It intelligently prioritizes semantic interaction via accessibility trees to minimize token usage while maintaining high reliability across diverse web technologies.
AIPex Browser Control Use Cases
- Automate multi-step browser workflows such as form filling, account creation, or data entry.
- Extract web content and save pages as markdown for RAG pipelines or documentation.
- Perform automated UI and UX testing, including accessibility checks and regression verification.
- Capture high-resolution screenshots of specific browser tabs for visual auditing.
- Manage complex, multi-tab research tasks that require switching between various web sources.
How AIPex Browser Control Works
- The AI agent starts the aipex-mcp-bridge using npx to establish a communication channel.
- The user connects the AIPex Chrome extension to the bridge via a local WebSocket (defaulting to port 9223).
- The agent identifies elements on the page by querying the accessibility tree to retrieve unique element IDs (UIDs).
- Actions such as clicking, typing, or scrolling are executed directly using these UIDs for maximum precision.
- If semantic navigation fails, the agent falls back to vision-based interaction using screenshots and pixel coordinates.
AIPex Browser Control Setup
To integrate this functionality into your Openclaw Skills workflow, ensure you have Node.js >= 18 and the AIPex Chrome extension installed.
1. Register the MCP Server
Add the following configuration to your agent (Cursor, Claude Desktop, or Windsurf):
{
"mcpServers": {
"aipex-browser": {
"command": "npx",
"args": ["-y", "aipex-mcp-bridge"]
}
}
}
2. Connect the Extension
Open the AIPex extension options in Chrome, navigate to the WebSocket Connection section, enter ws://localhost:9223, and click Connect.
AIPex Browser Control Data Schema & Taxonomy
The skill manages browser state through Tab IDs and utilizes UIDs for element-level targeting. The following table describes the data structures handled by this skill:
| Component | Data Type | Purpose |
|---|---|---|
| Tab ID | Integer | Identifies specific browser tabs for multi-tasking |
| Element UID | String | Unique identifier for interactive DOM elements |
| Page Metadata | JSON | Contains URL, title, and loading status |
| Screenshots | Base64/PNG | Visual representation of the tab state for vision models |
| Markdown | String | Extracted page content formatted for easy LLM consumption |
This structured approach ensures that Openclaw Skills can maintain context across complex browsing sessions.
AIPex Browser Control Advanced Features
- Smart element searching that utilizes the accessibility tree to find buttons and inputs without high-cost visual processing.
- Human-in-the-loop intervention tools that allow the agent to pause for manual user input during sensitive workflows.
- Comprehensive tab management including pinning, grouping, and cross-window navigation.
- Pixel-accurate computer control for interacting with canvas elements and non-standard web components.
- Direct-to-markdown content extraction to streamline data gathering for other Openclaw Skills.
SKILL.md
Loading
Related Openclaw Skills
TorrentClaw
TorrentClaw is a powerful AI agent skill for searching, filtering, and downloading high-quality movie and TV torrents with automated local client integration.
Reddit Readonly
A read-only Reddit browsing tool for AI agents that extracts posts and comments via public JSON endpoints without requiring write access.
SocialRails
A comprehensive social media management skill that enables post scheduling, AI caption generation, and analytics tracking via chat commands.
fdv.lol - Headless Memecoin Auto Trader
A CLI-based automated memecoin trader for Solana that leverages Agent Gary for autonomous AI-driven decision making.
Stock Copilot Pro
A professional-grade AI investment assistant providing deep multi-market stock analysis and real-time sentiment tracking.
ClawDNA
ClawDNA is a specialized utility that synthesizes historical AI agent interactions into structured, privacy-safe public personas and wiki-style profiles.
Featured*
