A high-speed, cost-effective headless browser for AI agents that utilizes accessibility trees for efficient web navigation and interaction.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install gbrow
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 gbrow 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 Gbrow?
Gbrow is a comprehensive headless browser skill powered by Playwright and Bun, specifically engineered to enhance the capabilities of AI agents. Unlike traditional methods that rely on expensive vision models to interpret screenshots, Gbrow uses the accessibility tree to provide a structured, text-based view of web pages. This makes it significantly faster and more reliable than vision-based alternatives, providing a robust foundation for building complex workflows within Openclaw Skills. By mapping interactive elements to unique reference tags, it allows agents to navigate, click, and fill forms with extreme precision without the overhead of processing high-resolution images.
Gbrow Use Cases
- High-speed web scraping and data extraction where traditional CSS selectors are too fragile.
- Automated form filling and account management on complex web applications.
- Building cost-effective web research agents that avoid expensive vision-model API calls.
- Testing web interface accessibility and interaction flows within Openclaw Skills automation.
How Gbrow Works
- The skill initializes a local server using Bun and Playwright to control a headless Chromium instance.
- When an agent visits a URL, Gbrow generates an accessibility tree snapshot (ariaSnapshot) rather than a visual screenshot.
- Every interactive element on the page is assigned a reference handle, such as @e1 or @e2, allowing for simple referencing.
- The AI agent interprets the structured text tree and issues commands like click, fill, or scroll using the specific element references.
- Commands are sent via an HTTP API to the Gbrow server, which executes the browser action in real-time and returns the result.
Gbrow Setup
To integrate Gbrow into your Openclaw Skills environment, follow these installation steps:
# Clone the repository into your skills directory
git clone https://github.com/ashish797/Gbrow.git ~/.openclaw/workspace/skills/Gbrow
# Navigate to the directory and run the setup script
cd ~/.openclaw/workspace/skills/Gbrow
bash setup.sh
# Start the browser server
bun run src/server.ts
Gbrow Data Schema & Taxonomy
Gbrow organizes its operational state and communication through structured JSON and text formats:
| Component | Detail |
|---|---|
| State Management | Port and auth tokens are stored in .gstack/browse.json |
| Page Snapshots | Elements are mapped using a @ref [type] "label" taxonomy |
| API Protocol | JSON-based POST requests for navigation and interaction commands |
| Visual Output | Supports standard PNG screenshots and PDF exports when requested |
Gbrow Advanced Features
- Advanced snapshot flags to filter for interactive elements or limit tree depth for faster processing.
- Full multi-tab support allowing agents to manage multiple concurrent browsing sessions.
- Direct JavaScript execution support for complex in-page computations and data retrieval.
- Responsive viewport configuration and PDF generation for reporting and document archival.
- Secure token-based HTTP API for programmatic control of the browser instance.
SKILL.md
Loading
Related Openclaw Skills
Alibaba Cloud Qwen Text Embedding
Generate high-dimensional text embeddings via Alibaba Cloud Model Studio for search and AI retrieval.
Smart Compact
A sophisticated 4-phase progressive compaction strategy for Openclaw Skills that safeguards critical information before context compression.
Alibaba Cloud Qwen Rerank
A specialized skill for reordering search candidates using Alibaba Cloud Model Studio rerank models to enhance retrieval relevance.
Alibaba Cloud Contact Center AI Management
A powerful utility for managing Alibaba Cloud Contact Center AI resources and automating OpenAPI interactions within Openclaw Skills.
Aliyun OpenSearch Vector Search
A high-performance skill for managing vector documents and executing HA/SQL searches via the Aliyun OpenSearch ha3engine SDK.
Alibaba Cloud Qwen Multimodal Embedding
Generate high-dimensional multimodal embeddings from text, images, and videos using Alibaba Cloud Model Studio.
Featured*
