BrowserOS enables AI agents to control a real Chromium browser for complex web interactions like filling forms, clicking elements, and managing browser states.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install browseros-agent
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 browseros-agent 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 BrowserOS?
BrowserOS is a sophisticated browser automation layer specifically designed for AI agents. As a key component among Openclaw Skills, it provides a bridge between LLMs and a live Chromium instance, allowing for precise interaction beyond simple page reading. It excels in scenarios where agents need to perform human-like actions such as clicking buttons, filling multi-step forms, and navigating complex single-page applications (SPAs).
Unlike traditional scrapers, BrowserOS maintains a stateful connection to the browser, offering a snap-act-resnap loop that ensures high reliability even as the DOM changes. It treats the browser as a complete operating system for the web, providing access to tab management, bookmarks, and history while maintaining strict safety defaults to protect user data and workspace integrity.
BrowserOS Use Cases
- Automating multi-step login flows and form submissions
- Conducting multi-tab research and cross-referencing data between websites
- Extracting structured data from dynamic websites using CSS selectors or text-based snapshots
- Visual verification of web application states through screenshots and PDFs
- Programmatic management of browser resources like history, bookmarks, and tab groups
How BrowserOS Works
- Launch & Connect: The agent initializes the browser environment using browseros-cli and establishes a connection to a real Chromium instance.
- Navigation: The agent opens a new tab or navigates an existing one to a target URL, generating a unique Page ID.
- Snapshotting (Snap): A snapshot of the page is taken to identify interactive elements, assigning temporary numeric IDs to buttons, inputs, and links.
- Action (Act): The agent performs actions using those numeric IDs (e.g., click 15, fill 10) to avoid the fragility of raw CSS selectors.
- Observation (Re-snap): After any interaction, the agent re-snaps the page to capture the updated DOM state and new element IDs.
- Cleanup: Once the task is complete, the agent closes the tabs and releases resources to maintain system performance.
BrowserOS Setup
Ensure you have Node.js installed, then execute the following commands to integrate this into your Openclaw Skills workflow:
# Install the CLI tool
npm install -g browseros-cli
# Install the BrowserOS application components
browseros-cli install
# Launch the browser environment
browseros-cli launch
# Automatically configure the connection
browseros-cli init --auto
# Verify the installation status
browseros-cli health
BrowserOS Data Schema & Taxonomy
BrowserOS organizes interaction data through a structured hierarchy of sessions and pages. The following table describes the primary data outputs:
| Data Type | Description | Format |
|---|---|---|
| Page State | Metadata about open tabs, including IDs and URLs | JSON / Table |
| Element Map | Interactive elements mapped to numeric IDs for agent interaction | List |
| Content | Full page content or scoped selector results | Markdown / Text |
| Visuals | Snapshots of the current viewport or full page | PNG / PDF |
| Resources | Browser history, bookmarks, and group metadata | Structured List |
BrowserOS Advanced Features
- Multi-tab orchestration: Target specific tabs using the -p flag for complex research tasks
- Hidden browser mode: Run operations in the background with the --hidden flag
- JavaScript Evaluation: Execute custom scripts via the eval command for advanced data extraction
- Resource Management: Full CRUD operations for browser bookmarks, history, and tab groups
- Low-token content extraction: Use the text command to get markdown-formatted page content, minimizing token usage compared to screenshots
SKILL.md
Loading
Related Openclaw Skills
Local Business Money Radar
A high-intelligence lead generation tool that ranks local businesses by revenue opportunity using automated health diagnostics and AI scoring.
Real Estate Lead Machine
A high-performance lead generation engine that scrapes property portals to identify motivated sellers and automate personalized AI outreach.
LinkedIn B2B Buying Signal Detector
An AI-powered sales intelligence tool that identifies high-intent B2B prospects by analyzing hiring patterns, growth metrics, and tech stack changes on LinkedIn.
Human Plan Manager
A structured framework for managing short-term goals and detailed execution plans using simple Markdown files.
GKE Security Hardening Tool
A specialized security configuration generator that automates GKE cluster hardening according to Center for Internet Security (CIS) Benchmark standards.
Oracle OKE Security Hardening Tool
A professional-grade API tool that automates the generation of CIS Benchmark-compliant security configurations for Oracle Kubernetes Engine.
Featured*
