A high-performance Rust-based headless browser automation CLI that enables AI agents to navigate, interact with, and extract data from web pages using structured commands.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agent-browser-jau771
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 agent-browser-jau771 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 robust and fast headless browser automation tool designed specifically for the needs of AI agents and developers. Built with Rust and featuring a Node.js fallback, it provides a powerful interface for programmatically navigating the web. By utilizing an innovative ref-based system for DOM elements, it simplifies complex interactions like clicking, typing, and form filling into easy-to-execute commands.
As a standout tool within the Openclaw Skills library, it bridges the gap between raw LLM logic and live web environments. It allows agents to perceive the web through structured snapshots (accessibility trees) and perform human-like actions with high precision. Whether you are building an automated researcher or a UI testing bot, Agent Browser offers the reliability and speed required for production-grade AI agent workflows.
Agent Browser Use Cases
- Automated Web Data Extraction: Scrape structured data from modern, JavaScript-heavy websites that require interaction.
- AI-Driven UI Testing: Programmatically test web interfaces by simulating user flows like login, checkout, and form submission.
- Automated Form Completion: Use AI agents to fill out complex multi-step forms across different domains.
- Web Interaction Monitoring: Record sessions and take snapshots to track changes in web page layouts or content over time.
- Session-Based Automation: Save and load authentication states to bypass repetitive login steps in automated workflows.
How Agent Browser Works
- Navigation: The agent initializes the browser and navigates to a specific URL using the open command.
- Snapshot Generation: The tool generates a snapshot of the page, typically focusing on interactive elements, and assigns them unique reference IDs (e.g., @e1, @e2).
- Element Interaction: The agent performs actions like click, fill, or hover by targeting these stable reference IDs.
- State Management: If the page content changes or navigation occurs, a new snapshot is taken to update the references.
- Observation & Extraction: The agent retrieves text, attributes, or screenshots to verify the success of its actions before closing the session.
Agent Browser Setup
Install the Agent Browser CLI globally using npm to begin integrating it into your Openclaw Skills workflow.
npm install -g agent-browser
agent-browser install
agent-browser install --with-deps
For users building from the edge or contributing, you can install 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 its interactions through a structured command-line interface that can return both human-readable text and machine-readable JSON.
| Data Type | Description | Format |
|---|---|---|
| Snapshots | A tree structure of interactive elements or accessibility nodes. | Text or JSON |
| Session State | Saved cookies and local storage for persistent authentication. | .json files |
| Media | Visual captures of the browser state. | .png, .pdf, .webm |
| Network | Logs of HTTP requests and intercepted routes. | JSON List |
Agent Browser Advanced Features
- Parallel Browser Sessions: Run multiple isolated browser instances simultaneously using the
--sessionflag. - Authentication Persistence: Use
state saveandstate loadto manage complex login sessions across different executions. - Network Interception: Intercept, block, or mock network requests to control how the agent perceives site data.
- Device Emulation: Simulate mobile devices, set custom geolocations, and toggle offline modes for varied environment testing.
- Trace & Recording: Start video recordings or generate Chrome DevTools Protocol traces for deep debugging of agent behavior.
SKILL.md
Loading
Related Openclaw Skills
PancakeSwap Collect Fees
A specialized tool for discovering, calculating, and generating deep links to collect liquidity provider fees across multiple PancakeSwap versions and chains.
Bitget Wallet Skill
An advanced cryptocurrency management skill for interacting with Bitget Wallet APIs to perform secure swaps, market analysis, and multi-chain wallet operations.
Directory Listing Ads
A specialized framework for managing and optimizing paid advertising campaigns across major AI directories and software marketplaces.
Xiaojia Skills
A powerful integration for calling JustAI openapi async endpoints to generate structured marketing content and manage complex agent workflows.
TopYappers Social Media Intelligence
TopYappers provides AI agents with real-time social media intelligence, influencer discovery across major platforms, and viral content tracking.
Obsidian-Ontology Sync
Bidirectional synchronization that transforms unstructured Obsidian markdown notes into a structured, queryable ontology graph.
Featured*
