A high-speed, Rust-powered headless browser CLI that enables AI agents to navigate web pages and interact with elements using structured commands.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install junxuan
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 junxuan 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 Agent Browser?
Agent Browser is a cutting-edge automation tool designed to bridge the gap between AI agents and the web. Built with a high-performance Rust core and a Node.js fallback, it provides a robust command-line interface for navigating websites, interacting with DOM elements, and capturing page states. As a standout feature in the Openclaw Skills ecosystem, it utilizes a unique referencing system (@refs) that simplifies complex web interactions into predictable, executable commands, making it ideal for autonomous agents requiring reliable browser control.
This tool goes beyond simple scraping by offering full interaction capabilities, including form filling, session management, and media capture. Whether you are building automated research workflows or complex UI testing suites, this addition to your Openclaw Skills toolkit ensures your agents can handle modern, JavaScript-heavy websites with precision and speed.
Agent Browser Use Cases
- Automating complex web-based workflows such as flight bookings or software registrations.
- Extracting structured data from dynamic, client-side rendered applications.
- Performing automated UI/UX testing and generating visual bug reports with screenshots and video.
- Managing authenticated sessions by saving and loading browser states across different Openclaw Skills tasks.
- Mocking network responses and intercepting API calls for development and debugging purposes.
How Agent Browser Works
- Launch the browser instance and navigate to a target URL using the open command.
- Generate a page snapshot to map the accessibility tree and assign stable @refs to interactive elements.
- Execute specific actions like click, fill, or scroll by targeting the generated references.
- Re-snapshot the page after any navigation or significant DOM change to maintain interaction accuracy.
- Capture outputs such as JSON data, screenshots, or session state files to be used in subsequent Openclaw Skills steps.
Agent Browser Setup
To integrate this tool into your Openclaw Skills environment, follow these installation steps:
# Recommended installation via npm
npm install -g agent-browser
agent-browser install
agent-browser install --with-deps
# Alternative: Build from source
git clone https://github.com/vercel-labs/agent-browser
cd agent-browser
pnpm install
pnpm build
agent-browser install
Agent Browser Data Schema & Taxonomy
Agent Browser organizes interaction data and page metadata into structured formats for easy consumption by AI agents:
| Data Type | Format | Description |
|---|---|---|
| Page Snapshot | JSON/Text | Accessibility tree with @ref pointers for interactive elements. |
| Session State | JSON | Exportable file containing cookies and localStorage data. |
| Media Output | PNG/PDF/WebM | Visual representations of the page or recorded interaction sessions. |
| Network Logs | JSON | Tracked requests, headers, and filtered API activity. |
| Element Info | JSON | Bounding box coordinates, attributes, and visibility states. |
Agent Browser Advanced Features
- Parallel browser sessions using isolated profiles for multi-tasking.
- Comprehensive network control including request routing, aborting, and response mocking.
- Full media suite capable of generating full-page screenshots, PDFs, and WebM video recordings.
- Advanced device emulation for testing geolocations, viewports, and user agents within Openclaw Skills.
- Deep debugging support via headed mode, Chrome DevTools Protocol (CDP) connection, and execution traces.
SKILL.md
Loading
Related Openclaw Skills
Obsidian Writer
A specialized skill for generating Obsidian-compatible Markdown notes and visual Canvas diagrams with automated Git synchronization.
botauth
A secure CLI-based skill for retrieving secrets and credentials from a desktop vault with human-in-the-loop approval.
Clawscan
Clawscan is a specialized security auditing tool designed to perform first-pass vulnerability checks and malicious hash detection for your deployment.
Claude Code Controller
The Claude Code Controller is a robust integration for Openclaw Skills that enables seamless management of the Anthropic Claude Code CLI for autonomous software engineering.
Paragraph: Web3 Blogging & Newsletter Automation
A powerful Openclaw Skills integration for Paragraph.com that enables automated Web3 blogging, onchain content management, and community tokenization.
Inkroam Topic Expert
An automated content ideation pipeline that scrapes hot trends, scores them for viral potential, and syncs data to Feishu and Telegram.
Featured*
