A headless browser automation CLI designed for AI agents to navigate, interact with, and extract data from websites seamlessly.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install openclaw-agent-browser
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 openclaw-agent-browser 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 specialized command-line tool that enables AI agents to perform complex web interactions through a controlled Chromium instance. As a key part of the ecosystem of Openclaw Skills, it provides a structured way to handle headless browsing tasks like clicking buttons, filling out forms, and capturing visual data without manual intervention.
By abstracting the complexities of browser drivers into a simple CLI, it allows developers to build robust workflows that can navigate dynamic web applications. The tool focuses on reliability and state management, ensuring that Openclaw Skills can maintain persistent sessions and handle the asynchronous nature of the modern web.
agent-browser Use Cases
- Automating repetitive web-based tasks like form submissions and account registrations.
- Scraping data from modern, JavaScript-heavy websites that require visual rendering.
- Capturing full-page screenshots or generating PDFs of web documents for archival.
- Testing web application flows and verifying UI elements programmatically.
- Monitoring specific web pages for content updates or visual changes over time.
How agent-browser Works
- Navigation: The agent starts by opening a specific URL to initialize the headless browser session.
- Discovery: Running a snapshot command identifies interactive elements and assigns them short references like @e1 or @e2.
- Interaction: The agent performs actions such as filling inputs, selecting options, or clicking buttons using these generated references.
- Synchronization: Commands like wait are used to ensure the page has reached a stable state or that network activity has idled.
- State Refresh: Because the DOM changes frequently, the agent re-snapshots the page to refresh element references after any interaction that triggers navigation or dynamic updates.
- Output: Final data is retrieved via text extraction, screenshots, or by saving the session state for future use.
agent-browser Setup
To integrate this capability into your Openclaw Skills, run the provided setup script to install the agent-browser CLI and the required Chromium engine.
scripts/setup.sh
Basic usage involves opening a site and taking a snapshot to see available interaction points:
agent-browser open https://example.com && agent-browser snapshot -i
agent-browser Data Schema & Taxonomy
The skill organizes web data and session information through a structured referencing system and JSON-based state files.
| Component | Description |
|---|---|
| Element References | Temporary handles (e.g., @e1) used to target specific DOM elements in the current view. |
| Session State | JSON files storing cookies and authentication data to persist logins across runs. |
| Page Artifacts | Output files including PNG screenshots, PDF exports, and raw text dumps. |
| Security Config | Environment variables that define domain allowlists and output size limits. |
agent-browser Advanced Features
- State persistence allowing agents to save and load authentication cookies to bypass login screens.
- Support for parallel sessions, enabling the management of multiple independent browser contexts simultaneously.
- Visual regression testing tools through built-in screenshot diffing capabilities.
- Security boundaries to prevent context flooding and restrict the agent to specific allowed domains.
- Advanced command chaining for optimized, low-latency web automation sequences.
SKILL.md
Loading
Related Openclaw Skills
Browser Automation v2
Enterprise-grade browser automation for OpenClaw featuring robust resource management, tab cleanup, and smart waiting mechanisms.
NZTA Traffic
Access real-time New Zealand state highway traffic conditions, road closures, and live camera feeds via the Waka Kotahi NZTA API.
Moltbook Authentic Engagement
A sophisticated engagement protocol for AI agents on Moltbook that prioritizes genuine voice and quality interactions over automated growth hacking.
Commit Analyzer
A diagnostic tool that analyzes git commit patterns to evaluate the operational health and productivity of autonomous AI agents.
Tenzing Moltbook Skill
Tenzing is an autonomous assistant designed to distill complex information, organize workspaces, and facilitate multi-agent collaboration.
Vector Robot Controller
A comprehensive toolset for programmatic control and voice integration of Anki Vector robots via the wire-pod server.
Featured*
