A comprehensive CLI tool designed for AI agents to automate web browsing, interactive testing, and data extraction via Chromium.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agb
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 agb 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 (AGB)?
The Agent Browser skill provides a robust interface for AI agents to navigate and interact with the web as a human would. By wrapping a headless Chromium instance into a powerful CLI, it enables agents to perform complex actions such as clicking elements, filling forms, and managing sessions with ease. This skill is a foundational piece for building sophisticated Openclaw Skills that require real-time web access and interaction.
Beyond simple navigation, it offers deep inspection capabilities through accessibility tree snapshots and semantic locators. It handles the heavy lifting of browser lifecycle management, including cookie persistence, network interception, and multi-tab orchestration, making it an essential tool for developers building autonomous web-capable agents.
Agent Browser (AGB) Use Cases
- Automated end-to-end web application testing and bug discovery.
- Scalable data extraction and web scraping from dynamic, JavaScript-heavy websites.
- Automating repetitive web-based administrative tasks and form submissions.
- Generating visual documentation via screenshots, PDFs, and video recordings.
- Managing authenticated sessions for research and data gathering behind login walls.
How Agent Browser (AGB) Works
- Initialize a browser session by navigating to a target URL or connecting to an existing CDP port.
- Capture an interactive snapshot of the page to identify elements and map them to stable references like @e1 or @e2.
- Send interaction commands (click, fill, hover, scroll) targeting specific references or semantic roles.
- Extract data from the DOM or capture the visual state through screenshots and video.
- Manage the browser lifecycle by closing sessions or saving state to JSON for future Openclaw Skills workflows.
Agent Browser (AGB) Setup
The Agent Browser and Chromium are pre-installed and ready for use in the default environment. To start a basic session, use:
agent-browser open https://example.com
agent-browser snapshot -i
agent-browser close
For persistent Openclaw Skills sessions, you can use the --profile flag to save login states or --session to isolate parallel tasks.
Agent Browser (AGB) Data Schema & Taxonomy
The skill organizes interaction data and output files using the following structures:
| Data Type | Description | Format |
|---|---|---|
| Refs | Temporary interactive element identifiers | @e[number] |
| Session State | Saved cookies and localStorage for persistence | JSON |
| Snapshots | Accessibility tree and element metadata | Markdown/JSON |
| Visuals | Screenshots, PDFs, and WebM video recordings | Binary Files |
| Network | Intercepted request and response metadata | JSON Logs |
Agent Browser (AGB) Advanced Features
- Parallel browser sessions with complete isolation using the --session flag.
- Real-time video recording and trace generation for debugging complex Openclaw Skills workflows.
- Advanced network routing to block, abort, or mock API responses during automation.
- Semantic element discovery using ARIA roles, labels, and placeholders for more resilient scripts.
- Native support for SOCKS5 and HTTP proxies to facilitate geo-distributed testing.
SKILL.md
Loading
Related Openclaw Skills
voipms-sms
A professional-grade integration for the Openclaw Skills ecosystem that enables programmatic SMS sending and retrieval via the VoIP.ms API.
YouTube Channel API Skill
A high-performance automated data extraction tool for harvesting structured YouTube channel information via the BrowserAct API.
Google Image API Automation Skill
A high-performance automation tool for extracting structured metadata, thumbnails, and source links from Google Images without CAPTCHA or IP restrictions.
News Aggregator Skill
A sophisticated intelligence tool that aggregates and analyzes real-time data from 8 global sources for comprehensive tech and finance briefings.
OpenAI Codex Operator
An advanced execution skill that enables AI agents to run OpenAI Codex CLI commands for automated code generation, refactoring, and debugging.
Gmail Cleaner
A powerful Python-based automation tool for bulk cleaning, organizing, and managing multiple Gmail accounts using the Gmail API.
Featured*
