A high-performance Rust-based headless browser CLI designed to give AI agents full control over web navigation and page interaction.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install chen-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 chen-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 robust, Rust-based headless browser automation CLI that bridges the gap between AI agents and the web. By utilizing a high-performance engine with a Node.js fallback, it allows developers to build complex workflows where agents can programmatically navigate pages, extract structured data, and perform human-like interactions. As a key component in the Openclaw Skills ecosystem, it provides a structured way for agents to see the web through accessibility trees and snapshots, making it significantly easier to automate tasks without manual DOM traversal.
The tool is designed for speed and reliability, offering a unique referencing system (@e1, @e2) that simplifies element interaction. Whether you are building an automated testing suite or a web-scraping agent, this Openclaw Skills integration ensures your agent has a reliable interface to handle everything from form submissions to complex state management and video recording.
Agent Browser Use Cases
- Automating complex web interactions within AI agent workflows.
- Extracting structured data and accessibility trees from dynamic web pages.
- Programmatically filling out and submitting web forms for data entry tasks.
- Performing end-to-end UI testing and capturing visual regression snapshots.
- Recording web sessions and generating PDFs or screenshots for automated reporting.
How Agent Browser Works
- Navigation: The agent initiates a session by opening a specific URL using the command-line interface.
- Analysis: The agent takes a snapshot of the page, which generates an interactive map of elements with stable reference IDs (e.g., @e1).
- Interaction: Based on the snapshot, the agent performs actions like clicking, typing, or scrolling using the assigned reference IDs.
- State Management: The skill allows for saving and loading session states, such as cookies and local storage, to maintain authentication across sessions.
- Verification: After interactions, the agent re-snapshots or waits for specific network/UI states to confirm the success of the operation.
Agent Browser Setup
Install the package globally via npm to integrate it into your Openclaw Skills environment:
npm install -g agent-browser
agent-browser install
agent-browser install --with-deps
Alternatively, you can build from source using pnpm:
git clone https://github.com/vercel-labs/agent-browser
cd agent-browser
pnpm install
pnpm build
agent-browser install
Agent Browser Data Schema & Taxonomy
Agent Browser organizes data through snapshots and session states. The primary data structures include:
| Component | Description |
|---|---|
| Snapshot | A structured tree (JSON or text) of interactive elements with @ref IDs. |
| State (JSON) | Persistent session data including cookies and localStorage for authentication. |
| Media Outputs | Screenshots (.png), PDFs (.pdf), and Video recordings (.webm). |
| Logs | Trace files (.zip) and console output for debugging interactions. |
Agent Browser Advanced Features
- Multi-session support for running parallel browser instances with isolated contexts.
- Network interception and routing to block or mock API requests during automation.
- Semantic locators (role, text, label) as an alternative to reference-based targeting.
- Built-in video recording and tracing for auditing AI agent behavior.
- Geolocation, viewport, and device emulation to test mobile-specific web interfaces.
- Chrome DevTools Protocol (CDP) support for connecting to existing browser instances.
SKILL.md
Loading
Related Openclaw Skills
Asimov's Laws
An ethical reference framework based on Asimov's Laws of Robotics to provide philosophical guidance for AI agent behavior.
ETermin Integration for AI Agents
A comprehensive Openclaw Skills integration for managing ETermin appointments, availability, and customer data via AI agents.
Gadget Skill
An expert AI skill providing smart home reviews, hardcore tech evaluations, and geek-level buying advice.
Calendar Query
A professional-grade CLI integration for managing cross-calendar availability and event tracking via the gog utility.
Quant Strategy Bundle
A comprehensive suite of verified quantitative trading strategy frameworks designed for stock market analysis, backtesting, and signal generation.
PoC Validator
An automated vulnerability verification probe designed to replay raw HTTP requests and analyze server responses for exploit signatures.
Featured*
