A high-performance Rust-based headless browser CLI designed to empower AI agents with seamless web navigation and interaction capabilities.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agent-browser-backup-20260407
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-backup-20260407 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 command-line interface that allows AI agents to interact with the web much like a human would. Built with Rust for speed and featuring a Node.js fallback, it provides a structured way to navigate pages, click elements, fill forms, and capture snapshots. As a key component for those utilizing Openclaw Skills, this tool translates complex web structures into manageable refs, making it easy for agents to perform precise actions without manual DOM traversal.
The skill excels at handling modern web complexities, including single-page applications, iframes, and authenticated sessions. By providing machine-readable outputs and session management, Agent Browser ensures that web-based workflows are reliable, fast, and easily integrated into broader AI-driven automation pipelines.
Agent Browser Use Cases
- Automating complex multi-step web interactions and form submissions.
- Extracting structured data from dynamic websites for research or analysis.
- Programmatically testing web user interfaces to ensure consistency.
- Capturing screenshots, PDFs, or video recordings of web sessions for documentation.
- Managing persistent authenticated states across different automation runs using Openclaw Skills.
How Agent Browser Works
- Initialize a session or open a specific URL using the navigation commands.
- Generate a snapshot of the page to identify interactive elements and their unique reference IDs.
- Execute interaction commands such as click or fill using the assigned reference IDs (e.g., @e1).
- Monitor page state changes or wait for specific network conditions or text to appear before proceeding.
- Extract required data, capture visual media, or save the current browser state for future use.
Agent Browser Setup
Install the package via npm for the best experience within your Openclaw Skills environment:
npm install -g agent-browser
agent-browser install
agent-browser install --with-deps
Alternatively, you can build from source by cloning the repository and using pnpm to build the project and install dependencies.
Agent Browser Data Schema & Taxonomy
Agent Browser utilizes a reference-based system to map DOM elements to short, actionable IDs. This structure is highly beneficial for Openclaw Skills users needing predictable data output.
| Data Type | Description |
|---|---|
| Snapshot Refs | Unique identifiers (e.g., @e1) assigned to interactive elements for simplified CLI interaction. |
| JSON Output | Machine-readable results for snapshots and data extraction enabled via the --json flag. |
| State Files | JSON objects containing cookies and localStorage to persist authenticated sessions. |
| Media Artifacts | Support for PNG screenshots, PDF document generation, and WEBM video recordings of browser activity. |
Agent Browser Advanced Features
- Multi-session support for running isolated parallel browser instances.
- Advanced network routing to intercept, block, or mock API requests and responses.
- Semantic locators allowing agents to find elements by role, label, or text content.
- Full state management to save and load authentication cookies and local storage.
- Deep debugging tools including headed mode, console log access, and execution tracing.
- Integration via Chrome DevTools Protocol (CDP) for external browser control.
SKILL.md
Loading
Related Openclaw Skills
Li Xiang Perspective v2.0
A strategic mental operating system for AI agents based on the product-first and efficiency-driven philosophy of Li Auto founder Li Xiang.
Playwright Test Generator
An AI-driven code generation engine that transforms natural language, HTML, or live URLs into production-ready Playwright test suites.
William Li (Li Bin) Thinking System v2.0
A sophisticated cognitive model that simulates the strategic decision-making and long-term vision of William Li, founder of NIO.
Ma Huateng (Pony Ma) Perspective v2
A sophisticated AI persona skill that simulates the strategic mindset, product philosophy, and decision-making logic of Tencent founder Ma Huateng.
Duru Obsidian KB
A markdown-first knowledge management skill that automates the transition from raw source material to a structured, interlinked Obsidian wiki.
Zhihu Hot Trends via UAPIPRO
A powerful script to fetch and analyze real-time trending topics from Zhihu via the UAPIPRO API.
Featured*
